Research Collection

Doctoral Thesis

Shaping ceramics in small scale - from microcomponents to gassensors

Author(s): Heule, Martin

Publication Date: 2003

Permanent Link: https://doi.org/10.3929/ethz-a-004485083

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

Diss. ETH No. 14992

Shaping Ceramics in Small Scale – from Microcomponents to Gas Sensors

A dissertation submitted to the

SWISS FEDERAL INSTITUTE OF TECHNOLOGY ZURICH

for the degree of

DOCTOR OF NATURAL SCIENCES

presented by

MARTIN HEULE

Dipl. Chem. ETH

born November 2, 1973

citizen of Zurich and Widnau (SG), Switzerland

accepted on the recommendation of

Prof. Dr. Ludwig J. Gauckler, examiner

Dr. Nicolae Barsan, co-examiner

Zurich, 2003

In loving memory of my mother, Margrit Heule-Fuchs1947-2001

5

Acknowledgements

I am very grateful to my advisor, Prof. Dr. Ludwig J. Gauckler, for the confi-dence he placed in me and his valuable guidance throughout the duration ofthe thesis. By sending us PhD students around the world early on in ourprojects, we have numerous opportunities to meet the important researchersin the field. In addition, I would like to thank Dr. Nicolae Barsan from theUniversity of Tübingen (Germany) for all his advice on gas sensing, his spon-taneous offer to perform joint experiments and for accepting to be co-exam-iner.

Without external technical support, this project would not have been pos-sible. I am very much indebted to Dr. Stefan Blunier and Prof. Dr. Jürg Dualfrom the Institute of Mechanical Systems IMES not only for allowing me touse their clean room equipment continuously, but also for giving me truemicrofabrication support. Dr. Daniel Bächi and Dr. Nicolas Szita taught mehow to operate in a clean room in the best possible manner. Membrane-basedmicrodevices would not work without silicon nitride layers of outstandingquality. These were coated for me by Dr. B. Ketterer and F. Glaus from PSIVilligen.

A very cordial ‘thank you’ goes to all my collegues, especially SibylleVuillemin and Urs Schönholzer, who together with me constituted the pow-erful micropatterning group at Nonmetallic Materials. I thank the peoplewho shared their office with me, Nicholas Grundy, Bengt Hallstedt andBeate Balzer, Lorenz Meier for extensive and refreshing discussions amongchemists, the people working on SOFC for sharing electrochemical know-how and equipment, Michael Jörger, Christoph Kleinlogel, Anja Bieberle,Eva Jud, Dainius Perednis, Michel Prestat, Brandon Bürgler, Jennifer Ruppfor her assistance concerning the Tübingen-measurements, Kurosch Rezwanfor helping in the enzyme-microreactor project, Hans Wyss for Hg-porosim-etry measurements and finally Peter Kocher for his high-precision workshoppieces.

6

Luana Cavalli was courageous enough to perform her diploma thesis explor-ing one of my fancy ideas. Thanks for pushing it through! I was also verylucky to have numerous undergraduate students choosing my topic for theirsemester projects: Michael Werner, Lukas Pfister, Pascal Jud, Roger Bach-mann, Julia Schell, Beatrice Sutter (all ETHZ), Lauren Ellery (CorpusChristi College, Cambridge, UK) and Nathan Yoder (Purdue University,West Lafayette, IN, USA).

In addition, there were many people supporting my work, I would like torecognise in particular:

• Roger Michel for continuous discussion and scientific exchange about soft lithography and help with ToF-SIMS measurements.

• Didier Falconnet, Prof. Dr. Marcus Textor, Laboratory of Surface Science and Technology, ETHZ Schlieren also for ToF-SIMS measurements.

• Prof. Dr. Nicholas Spencer, LSST, for allowing us to sneak constantly inside his laboratories in order to use the plasma cleaner and the contact angle measurement setup.

• Dr. Alexander Gurlo for spending almost one week of his precious time for joint measurements in Tübingen.

• Dr. Thomas Bürgi (Technical Chemistry) for access to their e-beam evap-oration machine with which early versions of the gas sensors were coated with Pt and SiO2.

• Hans-Ruedi Scherrer and his apprentices (Physics Dept Workshop) for coating some microhotplate wafers with excellent Pt-layers.

• Anna Mezzacasa and Prof. Dr. Ari Helenius (Institute of Biochemistry, ETHZ), for letting me alienate their DNA-injecting equipment for the selective doping of microceramics.

• Yves Kaufmann for helping me to correct this thesis.

My family have supported me in all my decisions and never failed to supportme in need. This is an excellent opportunity to say ‘thank you for all’!

Finally, I would like to thank Mirjam Holderegger for all her constantsupport and patience during the sometimes awkward writing process.

9

Contents

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Zusammenfassung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

General 1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.1. Miniaturisation Science and Technology. . . . . . . . . . . . . . . . . . . . . . . . . . . 171.2. Photolithography and IC industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181.3. Soft Lithography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211.4. Gas Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241.5. Metal Oxide Gas Sensors on Micromachined Substrates . . . . . . . . . . . . 331.6. Aim of the Study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381.7. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Overview on Powder-based Ceramic 2. Meso- and Microscale Fabrication Processes . . . . . . . . . . . . . . . . . . . . . . . 47

2.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472.2. Direct writing methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502.3. Downsizing Mechanical Processing Methods . . . . . . . . . . . . . . . . . . . . . . 542.4. Lithography-based Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572.5. Self-Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592.6. Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 612.7. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

Powder-Based Microcomponents on Silicon 3. Substrates fabricated by Micromolding in Capillaries . . . . . . . . . . . . . 69

Patterning Colloidal Suspensions by 4. Selective Wetting of Microcontact-Printed Surfaces . . . . . . . . . . . . . . . 85

Casting of Suspensions 5. into Photoresist Structures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

10 C O N T E N T S

Miniaturised Enzyme Reactor Based on 6. Hierarchically Shaped Porous Ceramic Microstruts . . . . . . . . . . . . . . 109

Gas Sensors Fabricated from Ceramic 7. Suspensions by Micromolding in Capillaries . . . . . . . . . . . . . . . . . . . . . 121

Miniaturised Tin Oxide Gas Sensors 8. on Microhotplates by Micromolding in Capillaries . . . . . . . . . . . . . . . 133

Increasing the Integration Density by Vertically9. Separating the Heater of the Microhotplate Design . . . . . . . . . . . . . 149

Validating the Concept of 10. Miniaturising Resistive-type Gas Sensors . . . . . . . . . . . . . . . . . . . . . . . . 157

Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16511.

Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171A.A.1. Microsystem Design Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171A.2. Improved Clean Room Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181A.3. Thermal Characterisation of Microhotplates . . . . . . . . . . . . . . . . . . . . . 189A.4. Powder Characteristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199A.5. ToF-SIMS Spectra of Microcontact-Printed Surfaces . . . . . . . . . . . . . . 204A.6. List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211A.7. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

Curriculum Vitae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

11

ummary

Thin film ceramic layers are an integral part of microsystems technology.However, colloidal dispersions as used in classical ceramic powder processinghave not been considered as building blocks for microdevices until recently.Ceramics have several advantages over other materials also in microsystems,e.g. heat resistance, hardness, corrosion resistivity or functional properties.The purpose of this interdisciplinary thesis is to integrate colloidal systemsinto microfabrication technology.

As a main result, the prototype fabrication of miniaturized semiconduct-ing gas sensor arrays based on tin oxide powders is demonstrated. A single gassensor is barely visible by the human eye and covers only 10 by 30 µm2

which is two orders of magnitude smaller than most of today’s microsensordesigns. The miniaturized sensors could be integrated on state-of-the-artmicrohotplate substrates based on silicon micromachining. The techniqueswhich make possible the direct use of a colloidal dispersion for generatingfunctional ceramic microstructures are derived from soft lithography andphotolithography. The backbone of the thesis is a series of articles. It is orga-nised as follows:

Chapter 1 contains general introductions to microfabrication technologiesas they are used in this work, soft lithography and the current state in thefield of solid state gas sensing devices. In Chapter 2, a review article dealswith processing techniques that possibly allow for the introduction of pow-der-based ceramic components into microelectromechanical systems. In thefirst experimental part in Chapter 3, techniques from soft lithography are

S

12 S U M M A R Y

adapted to colloidal systems. The technique which is mainly used is micro-molding in capillaries (MIMIC). Formed in polydimethylsiloxane elas-tomers, microchannels served as molds which then were spontaneously filledwith suspensions of 0.1 - 40% solids loading owing to capillary forces. Theresulting micro-thickfilm structures have a height of several micrometers andtherefore differ from usual ceramic thin film coatings. By adjusting only thesolid content of the suspensions, even smaller lines of 1 - 2 µm width couldeasily be prepared. The other important soft lithography scheme, microcon-tact printing (µCP), is discussed in Chapter 4. Micropatterns of ceramicpowders can be obtained by selective wetting of microcontact-printed sur-faces. Aqueous colloidal dispersions adhered only to the hydrophilic micro-patterns whereas they were repelled by the hydrophobic surroundings in asimple dip coating process. Printing and selective wetting were carried outsuccessfully on two different ink/substrate combinations with a resolution of5 µm. The third microfabrication approach is photoresist casting (PRC,Chapter 5). Ceramic microstructures with cross sectional areas of5 by 10 µm2 were obtained. By using low-cost mask lithography, the PRCprocedure could be tested in a microfluidic application, as presented inChapter 6. A microfluidic continuous flow enzyme reactor consisting of alu-minum oxide microstruts mounted inside PDMS channels was built. Owingto the microdesign of the struts, a two-fold hierarchical structure wasobtained. First, the placement of the struts enforces mixing due to the multi-ple splitting of the laminar flow and subdivides the main channel into many15 µm wide channels. Second, the struts consist of loosely sintered particlenetworks exhibiting a well defined porosity of 60 nm. Compared to a micro-channel system without these struts, a five-fold increase in enzymatic productformation was obtained.

The second part of the thesis beginning with Chapter 7 deals exclusivelywith gas sensing applications. Miniaturized tin oxide gas sensors on bulk sub-strates were fabricated and characterised first. The MIMIC procedure wasintegrated in a sequence of conventional photolithography steps. Miniatur-ized gas sensors on sapphire single crystal substrates showed responses for100 ppm of hydrogen and 600 ppm for carbon monoxide obtained from anactive sensing area of only 10 by 40 µm2. In Chapter 8, miniaturized gas sen-

S U M M A R Y 13

sors are fabricated on microhotplates which consume only mW heatingpower. Additionally, the concept of placing one gas sensor on one microhot-plate is extended towards a gas sensor array consisting of several sensors onone hotplate. Twelve miniaturised gas sensors of nanoparticulate tin oxidewere integrated as an array on a single microhotplate by using micromoldingin capillaries. The integration density of gas sensors on microhotplates waseven more increased, as discussed in Chapter 9. The heater structures wereburied in a more complex microfabrication scheme using 5 masks. Processingsteps to prepare a 20 sensor array on microhotplates are presented and dis-cussed with regard to processing sequence, sensitivity to 1000 to 1500 ppmhydrogen and power consumption. Finally, the concept of miniaturisingsemiconducting gas sensors could be tested and validated in collaborationwith the Institute of Physical and Theoretical Chemistry of the University ofTübingen (IPTC). Gas sensors were prepared with MIMIC-microlines ofIn2O3 powder on substrates from IPTC. It is possible to miniaturize semi-conducting gas sensors without affecting the basic sensitivity of the sensors.

Fig. 1. Ceramic microline structures on a microhotplate compared with a human hair lying across.

14 S U M M A R Y

Zusammenfassung

Dünne keramische Filme sind integrale Bestandteile in den Mikrofabrikati-onstechnologien. Hingegen wurden kolloidale Dispersionen, wie sie in derklassischen keramischen Pulververarbeitung verwendet werden, bisher nichtals mögliche Ausgangsmaterialien für Mikrosystemkomponenten erachtet.Keramik bietet auch in Mikrosystemen diverse Vorteile gegenüber anderenMaterialien, z.B. Hochtemperaturbeständigkeit, Härte, Korrosionsbeständig-keit oder diverse funktionelle Eigenschaften. Diese interdisziplinäre Arbeitbefasst sich mit der Integration von Kolloidalen Suspensionen in die Mikro-fabrikationstechnologie.

Das wichtigste Resultat dieser Arbeit besteht in der Entwicklung vonminiaturisierten Gassensor-Arrays, welche aus halbleitendem Zinnoxidpul-ver hergestellt wurden. Ein einzelner Gassensor ist kaum noch sichtbar fürdas menschliche Auge und benötigt nur eine Fläche von 10 mal 30 µm2, waszwei Grössenordnungen kleiner ist als die der meisten modernen Mikrosen-soren. Die miniaturisierten Sensoren wurden erfolgreich auf modernen,mikrosystemtechnisch hergestellten Microhotplates integriert. Soft lithogra-phy und Photolithographie sind Techniken, die die direkte Erzeugung vonMikrostrukturen aus kolloidalen Systemen erlauben. Das Rückgrat dieserArbeit bildet eine Reihe von Veröffentlichungen. Sie sind wie folgt organi-siert:

Kapitel 1 enthält generelle Einführungen zu Mikrofabrikationstechniken,wie sie verwendet wurden, zu Soft Lithography und zu halbleitenden Gassen-soren. Im nächsten Kapitel 2 werden in einem Übersichtsartikel alternative

S U M M A R Y 15

Prozesse besprochen, die pulver-basierte Komponenten in der Mikrosystem-technologie ermöglichen. Im ersten experimentellen Kapitel 3 wird Microm-olding in Capillaries (MIMIC) behandelt. In poly-dimethylsiloxan geformteMikrokanäle können spontan mit Hilfe der Kapillarkraft mit Suspensionengefüllt werden, welche Volumenanteile von 0.1 - 40% aufwiesen. Die resul-tierenden Mikrostrukturen haben eine Höhe von einigen µm und unterschei-den sich daher wesentlich von herkömmlichen keramischenDünnfilmbeschichtungen. Noch dünnere Linien von 1 - 2 µm Breite könnenmit denselben Mikrokanälen durch das Anpassen des Volumengehalts derSuspension erzeugt werden. Das andere wichtige Verfahren aus der SoftLithography ist eine einfache Stempeltechnik, das Microcontact Printing(Kapitel 4). Mikrostrukturen aus keramischem Pulver wurden durch selekti-ves Benetzen von gestempelten Oberflächen erzeugt. Eine wasserbasierte kol-loidale Dispersion kann nur hydrophile Mikrostrukturen benetzen, währendsie von hydrophoben Flächen abperlt. Der Stempelprozess mit anschliessen-dem Beschichten mit Suspension konnte mit zwei Tinte-Substrat Kombina-tionen erfolgreich mit einer Auflösung von 5 µm durchgeführt werden. Derdritte Ansatz zur Mikrostrukturierung ist das Füllen von Photolackstruktu-ren, Photoresist Casting genannt (Kapitel 5). Mikrolinien mit Querschnittenvon 5 mal 10 µm2 wurden erzeugt. Mit dieser Technik wird in Kapitel 6 eineerste Anwendung demonstriert. Ein Mikro-Enzymreaktor mit kleinstenStrukturen aus Aluminiumoxid in einem PDMS-Kanal wurde gebaut. DieserAufbau hat eine zweifache Hierarchische Struktur: Erstens sorgt die versetztePlatzierung der keramischen Strukturen für eine erzwungene Durchmi-schung der laminar fliessenden Substratlösung und unterteilte den Hauptka-nal in viele kleine, 5-15 µm breite Subkanäle. Zweitens bestehen dieMikroelemente aus locker gesinterten Partikeln, welche eine gut definiertePorengrösse von 60 nm aufweisen. Verglichen mit einem Kanalsystem ohnediese Mikrokeramiken, hat dieser Enzymreaktor eine fünffach höhereUmsatzrate.

Der zweite Teil der Arbeit befasst sich ausschliesslich mit der Anwendungdieser Mikrokomponenten als Gassensoren. In Kapitel 7 wird die Herstel-lung von miniaturisierten Zinnoxidsensoren auf Saphirsubstraten beschrie-ben. Mittels MIMIC wurde ein Zinnoxid Sensorarray hergestellt. MIMIC

16 S U M M A R Y

wurde in eine Sequenz von konventionellen Photolithographieschritten inte-griert. Die miniaturisierten Gassensoren mit einer aktiven Fläche von nur10 mal 40 µm2 zeigten deutliche Signale für 100 ppm Wasserstoff und für600 ppm Kohlenmonoxid. Im folgenden Kapitel 8 wurden die miniaturisier-ten Gassensoren auf mikrofabrizierten, heizbaren Substraten, sogenanntenMicrohotplates platziert. Zwölf miniaturisierte Gassensoren aus SnO2 Nano-pulver wurden mit MIMIC auf einem Microhotplate als Array integriert.Kapitel 9 zeigt, wie die Integrationsdichte von Gassensoren durch die verti-kale Trennung von Heizstrukturen und Messelektroden weiter erhöht werdenkann. In Zusammenarbeit mit dem Institut für Physikalische und Theoreti-sche Chemie der Universität Tübingen (IPTC) bot sich die Gelegenheit, dievorgestellten Miniaturisierungskonzepte zu validieren (Kapitel 10). Gassenso-ren mit MIMIC-Linien aus In2O3 wurden auf IPTC-eigenen Substraten her-gestellt und ausgemessen. Damit konnte gezeigt werden, dass diegassensitiven Eigenschaften trotz Miniaturisierung erhalten bleiben.

Abb. 1. Keramische Mikrolinien auf einem Microhotplate im Vergleich zu einem schräg darüberliegenden menschlichen Haar.

17

General Introduction

1.1. Miniaturisation Science and Technology

As miniaturisation of electronic circuits progresses, other disciplines followand new areas of application arise. Based on established microfabrication pro-cesses from the IC industry, systems generally called MicroelectromechanicalSystems (MEMS) have been developed.[1,2] MEMS are tiny sytems that arenot ‘only’ able to calculate or to store data but to move, which constitutes theclassical MEMS area, to sense and to manipulate light,[3,4] to send and toreceive,[5,6] to detect and to dispense chemicals in liquid[7] and gaseous form.As this list of abilities indicates, the course towards small and intelligentautonomous systems for a variety of purposes is set. Examples of more com-plex systems are a micromachined scanning confocal optical microscope[8] orthe Digital Micromirror Device (DMD) projection chip by Texas Instru-ments.[9,10] Up to 1.3 million movable micromirrors (1280 x 1024 pixel,SXGA format), each 16 µm square, are integrated on one CMOS chip. Bytilting each mirror individually, a light beam can be switched on and off.These DMD-chips are successfully used for projection display applications.

The sheer reduction of size is not the only rationale of device miniaturisa-tion. Other properties like volume (~ l3), surface (~ l2), diffusion (~ l0.5) ortime (~ l0) scale differently with respect to length scale l. Thus, one can make

1.

18 C H A P T E R 1

not only smaller but more powerful devices. To illustrate this, the micromo-tors fabricated by Sandia National Laboratories rotate at 500’000 rpm - a carengine rotates at a few thousand rpm.[11]

The fabrication methods of MEMS allow a massive parallel production ofmultiple devices in one processing step. Owing to energy-intensive cleaningsteps and expensive clean room facilities, MEMS processes in general aremuch more expensive than conventional manufacturing. However, the massproduction of MEMS devices can bring down the cost of a single microde-vice considerably, in certain cases down to only a few cents per device. Sincethe number of available microsystems is not a limiting factor, the distributionof MEMS functionality to the single user becomes possible. Medical analysisprocedures can be made considerably faster by using microfluidic chips, andphysicians are sometimes enabled to carry out an analysis themselves ratherthan having to send blood or tissue to medical laboratories. This develop-ment would not be possible without the use of MEMS technology.

1.2. Photolithography and IC industry

The basic process in microfabrication is photolithography. Its origins dateback to 1822 when Nicéphore Niépce transferred an image to a glass platecovered with a mixture of lavender oil and bitumen (asphalt).[12] He placedan oil-painted paper on top and exposed the plate to the sun. After a fewhours, the illuminated areas were hardened and the unexposed areas could bedissolved in terpentine/lavender oil solution. This was the first application ofa process later to be called ‘negative’ photolithography.

Today, photolithography is the basis of all electronic processes of the ICindustry. With some delay, photolithography has also been used for the fabri-cation of MEMS. For further reference, the book Fundamentals of Microfabri-cation - the Science of Miniaturisation by Marc Madou gives an excellentoverview on all MEMS-relevant processes.[13] It has been used for referencethroughout this work.MEMS and electronics, although based on almost identical processes, are vir-tually separated. The CMOS[14] (Complementary Metal Oxide Semiconduc-tor) and VLSI (Very Large Scale Integrated Circuit) technologies producingthe current generation of 90 nm chip layouts[15] are sophisticated to such an

I N T R O D U C T I O N 19

extent that no deviations from the predefined standard processing routes arepossible. Additionally, there is Moore’s law first stated in 1965 predicting theprogress of miniaturisation in electronics.[16] Moore claimed that it would bepossible to increase the integration density in ICs by a factor of two everyyear. So, the industry is struggling to keep up with that pace. Intel Corp.expects to produce 45 nm circuits in 2007.[17] The website of Intel providesa very good source for current information on the state-of-the-art in the ICindustry. Another series of review articles concerning the future of electronicshave been published in Nature.[18-21]

Whereas the IC industry has long switched to exposing photoresists withcontactless direct projection optics,[22] the MEMS community predomi-nantly uses photolithography with hard masks for exposure. Fig. 1–1 displaysthe basic photolithography processing as it was used also in this work. Posi-tive resists are normally based on novolak resin mixed with a diazoquinonephotosensitizer,[23-25] negative resists on crosslinking epoxide resins. It iscommercially available as SU-8.[26]

20 C H A P T E R 1

Fig. 1–1. Basic photolithography. After exposing a photoresist with UV-light through a patterned mask (Cr metal patterned on a glass plate), the image is developed by dissolving the unexposed resist areas (negative resist, not shown), respectively the exposed areas (positive resist). For further processing, there are two approaches. a) lift-off process: Additive process by depositing another layer of material on top, subsequent removal of the resist and thus releas-ing the patterned material. b) surface micromachining: subtrac-tive approach by dry or wet chemical etching in which the photoresist protects areas from the etchant. Furthermore, the underlying layer, e.g. SiO2 may serve as additional mask for deep etching into silicon, called bulk micromachining.

silicon

spin coat photoresist

Lift-off Etching

Surface/Bulk Micromachining

expose through patterned Cr/glassmask by UV illumination

Cr

a) additive b) subtractive

SiO2/Si3N4 or similar layer

I N T R O D U C T I O N 21

1.3. Soft Lithography

Soft lithography is the enabling micropatterning technology for the directshaping of colloidal dispersions in this work. G. M. Whitesides proposed softlithography as low cost alternative to standard photolithography as early as1993.[27-30] The principal element is an elastomeric, transparent siliconepolymer for microstructure transfer. A commercially available elastomer ispolydimethylsiloxane (PDMS) which was initially proposed by the inventors.It was also used throughout this study. Patterned PDMS molds are preparedby casting a prepolymer onto a master structure and then crosslinking it.Finally, it can be peeled off the master and becomes a negative replica of themaster topography (Fig. 1–2a). Masters can be obtained from various sourcessuch as conventional photoresist structures and they can be reused severaltimes without the necessity for clean room equipment. This is a very impor-tant issue, since it opens the door for microfabrication also to researcherswithout clean room access.

There were several variations of microstructure transfer presented (Fig. 1–2b) in a ground-breaking review in Angew. Chem.[31] The focus is on astamping method called microcontact printing (µCP). A PDMS mold is sim-ply used as a stamp to transfer ink molecules onto a surface. An enormousrange of ink molecules are applicable, which makes it a true cross-disciplinarymicropatterning technique. Applications range from transferring simpleSAM-forming molecules to catalysts[32] and proteins or living cells.[33,34]

In Micromolding in Capillaries (MIMIC), a PDMS mold is placed upsidedown on a substrate. The microstructures are cut open at the side, where adroplet of liquid can be applied for spontaneous filling of capillaries.[35] Thismethod was primarily aimed at patterning organic polymers, but can beextended to other liquids, provided that they wet the PDMS capillaries andhave a sufficiently low viscosity.

The other two approaches are less known. Microtransfer molding (µTM)is the analogue to gravure printing, where the recessed structures are firstfilled with an ink material which is then transferred to the substrate. A sol-vent-soaked PDMS mold is placed on a solid layer of soluble polymer in Sol-

22 C H A P T E R 1

vent Assisted Micromolding (SAMIM). The mold sinks down into theliquefied layer, thereby shaping the surface into a replica of the master by fill-ing the recessed microcavities.

Fig. 1–2. Soft lithography with its variations. a) preparation of PDMS molds by cross-linking PDMS prepolymer over a master structure. b) main procedures for microstructuring liquid materials.

Micromolding in Capillaries (MIMIC) Microcontact Printing (µCP)

Solvent-assisted Micromolding (SAMIM)

Micro-transfer molding (µTM)

b) pattern Transfer

a) PDMS mold preparation

Master structure Demold PDMSCure PDMS prepolymer

Si OCH3

CH3 n

I N T R O D U C T I O N 23

At this point, the properties of PDMS are introduced in order to demon-strate its versatility in soft lithography. Apart from the commercial PDMSformulation that has become a de facto standard in soft lithography (Sylgard184, Dow Corning), Schmid et al. prepared other blends using differentmolecular weights and different cross-linking molecules. They were thus ableto optimize the surface hardness and Young’s modulus for nanometer resolu-tion µCP.[36] Sylgard 184 was used throughout this work. The curing chem-istry and components of Sylgard 184 are shown in Fig. 1–3. The maincomponent is a prepolymer of 18.500 D with vinyl terminal groups. Ahydrosilane/dimethylsiloxane (684 D) copolymer is the cross-linking agent.The resulting elastomer has interesting properties. The most important is thespontaneous wetting of surfaces, i.e. the PDMS stamp spontaneously estab-lishes a conformal contact to the surface. The work of adhesion to glass andgold was determined to be 0.1 J/m2 and 0.5 J/m2 respectively.[37] This prop-erty allows to seal the microcapillaries in MIMIC effectively even on sub-strates exhibiting a certain roughness, which is key to many applications.After structure transfer, the molds can be easily removed in a similar mannerlike the well known Post-It notes. In addition, PDMS is biocompatible andpermeable for gases which allows for patterning of living cells. By oxygenplasma treatment, the surface can be functionalized with hydroxyl functions,rendering it hydrophilic or enabling the coupling of other molecules. Twoplasma treated molds can be joined irreversibly by just pressing themtogether. All these properties make it extremely versatile not only in softlithography but also as rapid prototyping material in microfluidic applica-tions: Ismagilov et al. for example prepared complex-shaped microfluidicvalves of PDMS.[38]

24 C H A P T E R 1

Fig. 1–3. Components of Sylgard 184 PDMS. Cross-linking is achieved by a Pt-catalyzed hydrosylilation reaction.

1.4. Gas Sensors

The term “sensor” is used very often, in everyday life as well as in science andtechnology. If a definition of “sensor” is sought, one can choose from a vari-ety of differing suggestions. One concise suggestion for a definition states:

“The sensor is the primary part of a measuring chain which converts theinput variable into a signal suitable for measurement.”[39]

Without including or ruling out too many technicalities, this definitionreflects the principal function of a sensor, namely to convert an input variablesuch as temperature, pressure or in the case of chemical sensors, the presenceof a substance into a measurable signal. Typically, this is an electrical signalwhich can be processed electronically. There is much discussion whetherother parts in that measurement chain like signal amplification or condition-ing etc. should also be considered as a part of the sensor or not. For referringto the central part physically responsible for signal conversion solely, the term“transducer” is also used frequently.[40]

There are a lot of complementary and competing gas sensor technologiesrelying on different physical/chemical principles. See the book of Madou andMorrison[41] or the Sensors series from VCH[42] for further reference. A briefsummary is given in Table 1–1.

PDMS prepolymer

Hydrosilane cross-linker

cat.

Si O Si

Pt

SiOSi

Cross-link formed by hydro-silylation reaction

+

O Si O

n = 250

Si O Si OH

SiOSi

55

Si O Si O SiOSi

nO

Si

O

55

I N T R O D U C T I O N 25

Table 1–1. Various devices suitable for gas sensing. Adapted from [42] and from other sources where indicated.

Class Gas Sensor Detection Principle

Solid State Sensors[41]

Semiconducting Surface-catalytic combus-tion/interaction changes semiconductor electronic states, resp. conductivity

Potentiometric Nernst-potential on an elec-trode system, one electrode interacts with analyte

ChemFET[43] Field effect transistors with a gate material interacting with the analyte. I,V-curves become chemically sensitive

Amperometric Diffusion limited current of an ionic conductor

Calorimetric Heat of analyte reaction, e.g. catalytic combustion

Optical Sensors[42] Optodes IR, UV-VIS absorption spec-troscopy

Fluorescence Fluorescence excitation or quenching

SPR, Surface Plasmon Resonance

Interaction of an analyte layer with an evanescent field

...

Mass Sensitive Devices[42]

Acoustic[44]

Quartz Microbalance QMB, etc.

SAW, Surface-acoustic waves or BAW (bulk acoustic waves) excited on a quartz or piezoelectric substrate change in phase/frequency upon ad- or absorption of analyte in a suitable sorption layer (e.g. metals, polymers)

MEMS sensors Micro- or nanocantilevers bend mechanically upon adsorption of analyte[45-48]

26 C H A P T E R 1

For this work, semiconducting metal oxide sensors for measuring reduciblegases were miniaturised using a novel combination of standard and innova-tive microfabrication processes. In 1962, Seiyama et al. first presented thediscovery that semiconducting thin films of ZnO changed their electricalresistance (signal) as a function of varying gas concentrations (input variable).This simple device was employed as a novel detector for gas chromatogra-phy.[51] The layer had to be heated to 200°C and higher for maximum sensi-tivity. In the subsequent research, the focus has been drawn on SnO2, whichproved to be one of the most sensitive, in the case of oxidizing gases down tothe sub-ppm range[52] and versatile semiconductors for gas sensing.[53] Othersemiconducting metal oxides have also been used as summarized in theexamples of Table 1–2. The list is not complete as there is an ongoing searchfor novel gas sensitive materials.[54]

The fields of possible practical applications becomes apparent by lookingat the analyte gases in Table 1–2. Gas leakage detection in homes and inindustrial facilities (H2, CH4, etc.), fire warning systems (CO), environmen-tal monitoring (NOx), in automotive applications: cabin air quality monitor-ing (VOC, hydrocarbons), just to mention the most important ones. Moredemanding applications are still a domain of standard methods in analyticalchemistry, e.g. gas chromatography and mass spectrometry.

Polymer-based[49,50]

QMB, mass sensors A polymeric sorption layer increases in mass when ana-lytes absorb

Calorimetric Heat of solution is measured when analytes absorb into polymer layer

Table 1–1. Various devices suitable for gas sensing. Adapted from [42] and from other sources where indicated.

Class Gas Sensor Detection Principle

I N T R O D U C T I O N 27

1.4.1. Tin Oxide as Gas Sensing Material

The electrical resistance of SnO2 reacts to reducing species, e.g. CO, H2,Hydrocarbons, Ethanol and some oxidising gases like NOx and O3. At firstglance, tin oxide seems to be extremely versatile. However, the main problemis selectivity. The reducing gases mentioned cause a decrease in electricalresistance, the oxidating ones an increasing resistance which makes it difficultto discern between different molecules with the same sign of electrochemicalpotential.[63] There are two main approaches for selectivity improvement: toexploit the temperature-dependence[64-67] of the sensitivity of different mol-ecules and to dope the tin oxide with different catalytically active metals,mostly Pd,[66] Pt, In, Cu,[68,69] Nb, Mn[70] or Si.[71] Again, these are only afew successful examples as there are many researchers more or less aimlesslytrying to find new additives for improving metal oxide gas sensing. Therehave also been examples of mixing different metal oxides, like mixing TiO2

Table 1–2. Examples of Metal oxides used in conductivity type gas sensors for different analyte gases.

Metal Oxide Analyte Gas(es) Reference

SnO2 CO, H2, CnH2n+2 [53]

WO3 NOx, O3 [55]

In2O3 CO (“selective”) [56]

CdO-Fe2O3 C2H5OH [57]

TiO2 CO and NO2 [58]

TiO2-WO3 CH3OH [59]

WO3-Bi2O3 NOx [60]

ZnGaO4 Hydrocarbon mixtures (liquid petroleum gas)

[61]

Al-Fe2O3 CO, CH4 [62]

28 C H A P T E R 1

and SnO2 for butane sensing.[72] The selectivity for small molecules can beenhanced by depositing a diffusion filter layer such as SiO2 on top of thesemiconductor.[73]

Another more recent strategy was to come to terms with the selectivityproblem of semiconducting gas sensors.[74-81] Biological olfactory systemsinspired the idea of combining the signals of several imperfect sensors. Inhuman and mammalian noses, there are thousands of taste buds with badselectivity to different odors, but the brain is capable to derive a specific odoridentification by “processing” the signals from all taste buds. The technologi-cal analogue is hence often called “electronic nose” and it is tried to mimickthe natural olfactory process, see Fig. 1–4. An array of sensors with differentfunctionality is employed and their data are processed by statistical mathe-matics[82] or neural networks that are used to predict partial pressures of gasmixtures.[83] Massart and Vanderginste wrote a mile-stone textbook aboutthese multivariate mathematical methods, also called chemometrics.[84] It isimportant to note that only those species can be classified by electronic nosesthat produce a useful signal in at least one sensor of the array. Therefore, thesensitivity range of all the sensors in the array defines the application range.Classification of odors and taste becomes possible which is important forfood industries. Additionally, electronic nose systems have to be trained andcalibrated, e.g. the array needs to have measured the head space of “fresh”odor from fruit before being able to classify fruit with unknown freshness.Currently, the systems are far behind the natural capabilities of e.g. a dog’snose. One of the pioneers in the field is Lundström who used arrays of MOS-FET (Metal Oxide Field Effect Transistors).[43,85] Hong et al. used a micro-machined 4-sensor array of metal oxide sensing layers for detecting CO andEtOH mixtures.[86] Complex resistance data were evaluated as well.[87,88]

There has also been a potentiometric approach presented by Reinhard, amultielectrode setup on a ZrO2 ionic conductor.[89]

I N T R O D U C T I O N 29

Fig. 1–4. Electronic noses as a biomimetic approach.

1.4.2. Mechanisms of Gas Sensitivity

An understanding of the gas sensing mechanism on the molecular level isextremely difficult to obtain. There are few analytical methods that offer suf-ficient sensitivity for identifying molecules on hot surfaces or methods thatoffer the time resolution to capture reaction intermediates and kineticdata.[90] Under real sensing conditions there is already a complex gas mix-ture, air, humidity and the analyte gas. Therefore, most measurements weredone under better defined laboratory conditions, but lack to include the vari-ous additional effects from water. Therefore, the current mechanistic picturestill is very sketchy and lacks portability to real world gas sensors.

Multiple olfactory cells each withlimited capability

Array of sensorelements

Neural preprocessing

Data storage, matrix representation, statisticalrepresentation

Brain: derives overall impressionof smell / taste.

Pattern recognitionPCANeuronal Network

Highlighted cut through a tongue'staste bud.("array" of several receptor cells) Neural tissue

30 C H A P T E R 1

Fig. 1–5. Schematic of molecular species on a tin oxide surface.

Tin oxide is a semiconductor with a band-gap of 3.6 eV which makes italmost insulating at room temperature. Its semiconducting properties arebrought about by oxygen vacancies. As schematically shown in Fig. 1–5, tinoxide crystallizes in the rutile structure. The species noted in print are theprincipal ones that have been characterised spectroscopically among other,more complicated molecular assemblies by IR and EPR (electron paramag-netic resonance) and by TPD (temperature programmed desorption).[91] Arecent feature article by Barsan and Weimar summarizes the current state ofknowledge.[92] Tin oxide activates atmospheric oxygen, incorporating it intoits surface by reduction and building up a negatively charged layer. This equi-librium begins to establish above 100°C.

(1)

H2O

OHOHH O2- O O2-2- O O- +

CO CO2

wateractivation

carbon monoxide sensing

Sn4+

bulkoxygen

oxygendepletionlayer

β/2 O2,g + α e- + S O-αβS

I N T R O D U C T I O N 31

where α can take values of 1 or 2 depending on the oxygen reduction state,one or two-fold. β also may have values of 1 for single atom/ionic form and 2for molecular forms of oxygen. S denotes a free surface site bridging two tinions and g stands for gaseous. The electron(s) have to cross the negativelycharged barrier at the surface layer to reduce the oxygen molecule. In theenergy band structure diagram, this depletion layer leads to a bending ofstates towards the tin oxide surface (see Fig. 1–6). This is also a trapping ofelectrons and hence the electrical resistance of the topmost layer is increased.The situation is further complicated by the fact that water is similarly acti-vated (the resistance of tin oxide in dry atmosphere is lower than in wet con-ditions). By measuring the work function of the tin oxide, the effects of COand water could be characterised independently[93]

(2)

Indices Sn and O indicate ionic sites of the elements in the rutile lattice.There are at least two different species formed, a “hydroxylated” Sn-ion and a“protonated” oxygen. There may be an additional mechanism consuming alattice oxygen and creating an oxygen vacancy, two hydroxylated tin speciesand two electrons instead.

(3)

Processes like this presumably are also involved in ageing and unwanted sig-nal drift effects over longer times.[94,95]

An analyte molecule like CO reacts with the chemisorbed oxygen orhydroxyl groups and forms CO2 and electrons that are injected into the con-duction band. Macroscopically, a decrease in electrical resistance is measured.

(4)

H2Og + SnSn + OO (Sn+Sn - OH-) + (OH)+

O + e-

H2Og + 2 SnSn + OO 2 (Sn+Sn - OH-) + V++

O + 2 e-

β COg + O-aβS β CO2,g + α e- + S

32 C H A P T E R 1

IR absorption experiments reveiled the existence of intermediate carbonatespecies, coordinated in both unidentate or bidentate form.

Subsequently, one has to take into account the effects on grain size andfeatures like porosity. In powder-based sensors, it is estimated that the mostgas sensitive resistance lies in the necks, where particles are connected. Necksoften are very small, in the range of a few atomic layers. By model calcula-tions, different power laws for the electrical conductivity for different config-urations were obtained. However, the results are difficult to compare withreal gas sensing, since most sensors are doped with transition metals. Pd or Ptinduce a high density of new surface states which interact with the atmo-sphere similar to the properties discussed above. It was shown that Pd and Ptboth are oxidized at the tin oxide surface.[96] Among the analytical tech-niques used were XPS (x-ray photoelectron spectroscopy), Raman spectros-copy and TEM (transmission electron microscopy).[97] Before these results,most people argued that the catalytic enhancing of sensitivity or selectivityrelied on the catalytic activation of the analyte molecule by a metallic clusterwhich then diffuses towards the semiconductor for reaction with the oxygenlayer. This diffusion effect is called spill-over effect. It has been directlyobserved by Bennet et al. using STM microscopy.[98]

Additional effects may be induced by the electrode setup[99] for measur-ing the resistance which are difficult to characterise,[100] the contact resis-tance or other factors. In conclusion, one can not accurately predict neitherwhich sensitive material nor which sensor setup would be the best for a givenapplication.

1.4.3. Nanoscaled Powders Improve Sensitivity

A significant transition is encountered when using tin oxide nanopowders. Asthe radius r of nanoparticles approaches the depletion layer depth z0, whichmight be a few atomic layers (see Fig. 1–6), the higher band energy of thesurface charge layer becomes dominant throughout the material. Theobserved resistance is much more dependent on the state of the surface layer,i.e. dependent on the atmosphere. Additionally, the higher specific surfacearea should increase the sensitivity as well.

I N T R O D U C T I O N 33

Fig. 1–6. Energy band diagram for tin oxide surfaces.

As a matter of fact, nanoscaled powders were indeed found to have a largebeneficial impact on gas sensing.[101-104] Whether this model explanation issufficient for explaining all effects from the use of nanoscaled powders andthin films, is largely unknown.

1.4.4. Classic Sensor Designs

For standard resistive-type gas sensors, metal oxides are prepared as sinteredpellets that have been contacted with Pt wire. Resistance measurement andheating are done using the same electrodes. The predominant design consistsof a thin ceramic substrate onto which heating wires and sensing wires ofnoble metals have been screen-printed. This is the case in so-called Figarotype sensors (Figaro Engineering, Inc., Tokyo, Japan). Whereas they arecheap to produce, they have a significant disadvantage. The power consump-tion to maintain the substrate and sensing layer at temperatures of 200-800°C is rather high, typically a few watts. If the power should be cut to saveenergy, the times to heat up and stabilize the signal are prohibitingly low. Thekey for overcoming these problems is miniaturisation.

1.5. Metal Oxide Gas Sensors on Micromachined Substrates

The main problem to integrate metal oxide based gas sensors with microma-chined substrates are the high operating temperatures. Typical temperaturesin the sensing layers are up to 500°C which have to be maintained. Semanciket al. solved this problem in 1993 by embedding thin film heating filaments

E

depth from surface z0

Ef

Econduction band

Evalence band

z0

3.6 eV

34 C H A P T E R 1

on a freestanding dielectric membrane and introduced the term microhot-plates.[105] Their design based on a standard CMOS process which could beimportant for integration with electronics and for a cost-effective mass pro-duction owing to the wide availability of CMOS processes. How far the inte-gration of sensor and electronics has progressed, is impressively demonstratedby Hagleitner et al.[106] Their CMOS-based chip features an array of threedifferent sensors, a capacitive, a calorimetric and a cantilever gas sensor, tem-perature controller, preamplification, digitalisation of the signals and com-munication via a standard protocol.

The concept for lowering the heating power by orders of magnitude wasto minimize the mass to be heated and to thermally insulate it as effectively aspossible from the surrounding frame that is to remain at ambient tempera-ture. As membrane materials, thin film Si3N4, SiO2,[107] thin Si by CMOS-etch stop processes, even amorphous SiC films[108] or a combination of theseare used. Even 250 nm thin Si3N4 membranes exhibit astonishing mechani-cal strength and a low mass that gives the devices robustness even whenexposed to hard mechanical shocks. For the embedded heating filamentstructures and measuring electrodes, Pt or polysilicon are popular. In moreadvanced designs, there are also interlayers with good thermal conductivity,e.g. Al or a Si plug on the backside of the membrane to obtain a uniformtemperature distribution over the whole microhotplate area.[109] Integratedtemperature measurement was also used to determine the changes in temper-ature of -0.1°C to -4°C of the oxide layer upon gas exposure.[110] A compari-son of a standard classic design with the microhotplate developed within thiswork is shown in Table 1–3, and measured power consumption data aregiven in Fig. 1–7.

I N T R O D U C T I O N 35

The metal oxide layer is typically sputtered or evaporated since these are pro-cesses compatible to standard MEMS and CMOS processes.[111-113] But alsopowders have been deposited, either by droplet coating[107] or by screenprinting.[114] The layers are annealed on the chip by the integrated heater inall the referenced examples. Alternatively, the use of laser annealing was sug-gested by Steiner et al.[115]

Table 1–3. Comparison of a classic sensor design versus microhotplate sensors.

Classic Microhotplate

Steinel Hydrogen Sensor

12 Sensor Array M. Heule

Sensing Material Ga2O3 SnO2

Size of heated element

2.1 · 1.3 · 0.7 mm3 0.9 · 0.9 mm2 , 250 nm

Mass /mg 12.3 0.016

Thermal Mass /JK-1 5.3 · 10-3 3.6 · 10-6

Heating Time constant /s

5-10 0.005

Power consumption /W

Min. 0.6 Max. 1

Min. 0.01Max. 0.10

36 C H A P T E R 1

Fig. 1–7. Temperatures as a function of power consumption from bulk-sub-strate sensors and microhotplates.

In terms of micromachining, there are two approaches to generate the free-standing membranes (see Fig. 1–8). The first approach requires a photoli-thography step on the backside of the wafer to define the membraneopenings (Fig. 1–8a). Subsequently, the silicon wafer is etched until themembrane layer on the frontside is released. This etching process through the

800

700

600

500

400

300

200

100

0

Tem

per

atu

re /˚

C

2.52.01.51.00.50.0Power /W

Microhotplate no. 1 Microhotplate no. 2 Sapphire Microheater

fitted curves, T = a pb

800

700

600

500

400

300

200

100

0

Tem

per

atu

re /˚

C

0.001 0.01 0.1 1power /W

I N T R O D U C T I O N 37

whole wafer thickness is usually done using KOH or using deep reactive ionetching (DRIE). In the second approach, the membrane layers themselves arepatterned (Fig. 1–8b). Silicon is underetched through the resulting openingsuntil the membrane is released. In that design, the membrane is suspendedonly by four thin beams.

Extensive thermal characterisation data of the microhotplates fabricatedwithin this project are discussed in Appendix A3.

Fig. 1–8. Two microfabrication schemes to generate freestanding mem-branes for microhotplates. a) backside etching through the whole wafer thickness. b) frontside underetching through openings in the membrane layer. c) microhotplate chips. d) microhotplate glowing red hot at approx. 600 mW.

Side

Top

a b

38 C H A P T E R 1

1.6. Aim of the Study

The goal of this thesis is to explore processing techniques for the fabricationof miniaturized devices featuring tiny ceramic structures with dimensions inthe µm range. Materials and methods from microfabrication disciplinesshould be combined with classical colloidal powder processing of ceramics.As example for possible microdevice applications, the fabrication of a minia-turized array of semiconducting gas sensors was chosen.

I N T R O D U C T I O N 39

1.7. References

[1] S. D. Senturia, "Microsystem design", Kluwer Academic, Boston, 2001.[2] A. E. Franke, T.-J. King, R. T. Howe, Integrated MEMS Technologies, MRS Bulle-

tin, 2001, 26, 4, 291-295.[3] C. R. Giles, D. Bishop, V. Aksyuk, MEMS for Light-Wave Networks, MRS Bulle-

tin, 2001, 26, 4, 328-332.[4] N. Clark, A "Silicon Eye" Using MEMS Micromirrors, MRS Bulletin, 2001, 26, 4,

320-324.[5] C. T.-C. Nguyen, L. P. B. Katehi, G. M. Rebeiz, Micromachined Devices for Wire-

less Communications, Proc. IEEE, 1998, 86, 8, 1756-1768.[6] D. J. Young, Micromachining for rf Communications, MRS Bulletin, 2001, 26, 4,

331-332.[7] D. S. Peterson, T. Rohr, F. Svec, J. M. Frechet, Enzymatic Microreactor-on-a-Chip:

Protein Mapping Using Trypsin Immobilized on Porous Polymer Monoliths Molded in Channels of Microfluidic Devices, Anal. Chem., 2002, 74, 4081-4088.

[8] D. L. Dickensheets, G. S. Kino, Silicon Micromachined Scanning Confocal Opti-cal Microscope, IEEE J. MEMS, 1998, 7, 1, 38-47.

[9] L. J. Hornbeck, The DMD Projection Display Chip: A MEMS Based Technology, MRS Bulletin, 2001, 26, 4, 325-327.

[10] P. F. Van Kessel, L. J. Hornbeck, R. E. Meier, M. R. Douglass, A MEMS based Pro-jection Display, Proc. IEEE, 1998, 86, 8, 1687-1704.

[11] Sandia National Laboratories, Silicon Micromachines and other MEMS devices, http://mems.sandia.gov, 2002.

[12] T. W. Harris, "Chemical Milling", Clarendon Press, Oxford, 1976.[13] M. J. Madou, "Fundamentals of Microfabrication", 2nd ed., CRC Press, Boca

Raton, 2002.[14] U. Hilleringmann, "Mikrosystemtechnik auf Silizium", B. G. Teubner, Stuttgart,

1995, 10.[15] Intel Corp., M. Bohr, Presentation at the Intel Developer Forum, Sept 12,

http://www.intel.com/research/silicon/Bohr_IDF_0902.pdf, 2002.[16] G. E. Moore, Cramming more components onto integrated circuits, Electronics,

1965, 38, 8.[17] Intel Corp., Microarchitecture and Circuits Website, http://www.intel.com/labs/

microarch, 2002.[18] T. Ito, S. Okazaki, Pushing the Limits of lithography, Nature, 2000, 406, 1027-

1031.[19] A. I. Kingon, J. P. Maria, S. K. Streiffer, Alternative dielectrics to silicon dioxide for

memory and logic devices, Nature, 2000, 406, 1032-1038.[20] S. Loyd, Ultimate physical limits to computation, Nature, 2000, 406, 1047-1054.[21] M. H. Devoret, R. J. Schoelkopf, Amplifying quantum signals with the single-elec-

tron transistor, Nature, 2000, 406, 1039-1046.[22] R. F. Pease, Imprints offer Moore, Nature, 2002, 417, 6891, 802-803.

40 C H A P T E R 1

[23] A. Reiser, H. Y. Shih, T. F. Yeh, J. P. Huang, Novolak-Diazoquinone Resists: The Imaging Systems of the Computer Chip, Angew. Chem. Int. Ed., 1996, 35, 2428-2440.

[24] A. Reiser, Z. L. Yan, Y. K. Han, M. S. Kim, Navolak-diazonaphthoquinone resists: The central role of phenolic strings, J. Vac. Sci. Technol. B, 2000, 18, 3, 1288-1293.

[25] A. Reiser, J. P. Huang, X. He, T. F. Yeh, S. Jha, H. Y. Shih, M. S. Kim, Y. K. Han, K. Yan, The molecular mechanism of novolak-diazonaphthoquinone resists, Euro-pean Polymer Journal, 2002, 38, 619-629.

[26] H. Lorenz, M. Despont, N. Fahri, J. Brugger, P. Vettiger, P. Renaud, High aspect-ratio, ultrathick, negative-tone-near-UV-photoresist and its applications for MEMS, Sens. Actuators A, 1998, 64, 33-39.

[27] A. Kumar, G. M. Whitesides, Features of Gold Having Micrometer to Centimeter Dimensions Can Be Formed through a Combination of Stamping with an Elasto-meric Stamp and an Alkanethiol Ink Followed by Chemical Etching, Appl. Phys. Lett., 1993, 63, 14, 2002-2004.

[28] J. L. Wilbur, A. Kumar, E. Kim, G. M. Whitesides, Microfabrication by Microcon-tact Printing of Self-Assembled Monolayers, Adv. Mater., 1994, 6, 7-8, 600-604.

[29] R. J. Jackman, J. L. Wilbur, G. M. Whitesides, Fabrication of Submicron Features on Curved Substrates by Microcontact Printing, Abstr. Pap. Am. Chem. Soc., 1995, 210, 212.

[30] R. J. Jackman, J. L. Wilbur, G. M. Whitesides, Fabrication of Submicrometer Fea-tures on Curved Substrates by Microcontact Printing, Science, 1995, 269, 5224, 664-666.

[31] Y. Xia, G. M. Whitesides, Soft Lithography, Angew. Chem. Int. Ed., 1998, 37, 550-575.

[32] P. C. Hidber, W. Helbig, E. Kim, G. M. Whitesides, Microcontact printing of pal-ladium colloids: Micron-scale patterning by electroless deposition of copper, Lang-muir, 1996, 12, 5, 1375-1380.

[33] R. S. Kane, S. Takayama, E. Ostuni, D. E. Ingber, G. M. Whitesides, Patterning proteins and cells using soft lithography, Biomaterials, 1999, 20, 23-24, 2363-2376.

[34] G. M. Whitesides, E. Ostuni, S. Takayama, X. Y. Jiang, D. E. Ingber, Soft lithogra-phy in biology and biochemistry, Annu. Rev. Biomed. Eng., 2001, 3, 335-373.

[35] E. Kim, Y. N. Xia, G. M. Whitesides, Polymer Microstructures Formed by Mold-ing in Capillaries, Nature, 1995, 376, 581-584.

[36] H. Schmid, B. Michel, Siloxane Polymers for High-Resolution, High-Accuracy Soft Lithography, Macromolecules, 2000, 33, 8, 3042-3049.

[37] A. Bietsch, B. Michel, Conformal contact and pattern stability of stamps used for soft lithography, J. Appl. Phys., 2000, 88, 4310-4318.

[38] R. F. Ismagilov, D. Rosmarin, P. J. A. Kenis, D. T. Chiu, W. Zhang, H. A. Stone, G. M. Whitesides, Pressure-Driven Laminar Flow in Tangential Microchannels: an Elastomeric Microfluidic Switch, Anal. Chem., 2001, 73, 4682-4687.

[39] International Electrotechnical Committee, Terms and Definitions in Industrial Process Measurement and Control, IEC draft 65/84, 1982.

I N T R O D U C T I O N 41

[40] W. Göpel, K. D. Schierbaum, "Chemical and Biochemical Sensors", edited by W. Göpel, J. Hesse, J. N. Zemel, Sensors - A Comprehensive Survey, VCH, Wein-heim, 1991, 2.

[41] M. Madou, S. R. Morrison, "Chemical Sensing with Solid State Devices", Aca-demic Press, Boston, 1989.

[42] "Chemical and Biochemical Sensors", edited by W. Göpel, J. Hesse, J. N. Zemel, Sensors - A Comprehensive Survey, VCH, Weinheim, 1991.

[43] T. Eklov, I. Lundstrom, Distributed Sensor System for Quantification of Individual Components in a Multiple Gas Mixture, Anal. Chem., 1999, 71, 3544-3550.

[44] A. Hierlemann, A. J. Ricco, K. Bodenhöfer, W. Göpel, Effective Use of Molecular Recognition in Gas Sensing: Results from Acoustic Wave and in Situ FT-IR Mea-surements, Anal. Chem, 1999, 71, 3022-3035.

[45] C. B. Freidhoff, R. M. Young, S. Sriram, T. T. Braggins, T. W. O'Keefe, J. D. Adam, H. C. Nathanson, R. R. A. Syms, T. J. Tate, M. M. Ahmad, S. Taylor, J. Tunstall, Chemical sensing using nonoptical microelectromechanical systems, J. Vac. Sci. Technol. A, 1999, 17, 4, 2300-2307.

[46] M. K. Baller, H. P. Lang, J. Fritz, C. Gerber, J. K. Gimzewski, U. Drechsler, H. Rothuizen, M. Despont, P. Vettiger, F. M. Battiston, J. P. Ramseyer, P. Fornaro, E. Meyer, H. J. Guntherodt, A cantilever array-based artificial nose, Ultramicroscopy, 2000, 82, 1-4, 1-9.

[47] F. M. Battiston, J. P. Ramseyer, H. P. Lang, M. K. Baller, C. Gerber, J. K. Gimze-wski, E. Meyer, H. J. Guntherodt, A chemical sensor based on a microfabricated cantilever array with simultaneous resonance-frequency and bending readout, Sens. Actuator B, 2001, 77, 1-2, 122-131.

[48] J. Fritz, M. K. Baller, H. P. Lang, H. Rothuizen, P. Vettiger, E. Meyer, H. J. Guntherodt, C. Gerber, J. K. Gimzewski, Translating biomolecular recognition into nanomechanics, Science, 2000, 288, 5464, 316-318.

[49] H. N. McMurray, J. Albadran, Colorimetric and Fluorimetric Polymer Membrane Gas-Sensing Materials, MRS Bulletin, 1999, 24, 6, 55-59.

[50] M. Cole, J. W. Gardner, A. W. Y. Lim, P. K. Scivier, J. E. Brignell, Polymeric resis-tive bridge gas sensor array driven by a standard cell CMOS current drive chip, Sens. Actuators B, 1999, 58, 518-525.

[51] T. Seiyama, A. Kato, K. Fujiishi, M. Nagatani, A New Detector for Gaseous Com-ponents Using Semiconductive Thin Films, Anal. Chem., 1962, 34, 1502-1503.

[52] C. Cobianu, C. Savaniu, A. Arnautu, R. Iorgulescu, D. Dascalu, G. Leo, M. Mazzer, R. Rella, P. Siciliano, S. Capone, L. Vasanelli, A SnO2 microsensor device for sub-ppm NO2 detection, Sens. Actuators B, 1999, 58, 552-555.

[53] K. Ihokura, J. Watson, "The Stannic Oxide Gas Sensor", CRC Press, New York, 1994.

[54] J. Watson, K. Ihokura, Gas-Sensing Materials, MRS Bulletin, 1999, 24, 6, 14-17.[55] W. Zhang, H. Uchida, T. Katsube, T. Nakatsubo, Y. Nishioka, A novel semicon-

ductor NO gas sensor operating at room temperature, Sens. Actuators B, 1998, 49, 58-62.

[56] H. Yamaura, J. Tamaki, K. Moriya, N. Miura, N. Yamazoe, Selective CO Detec-tion by Using Indium Oxide-based semiconductor gas sensor, J. Electrochem. Soc., 1996, 143, 2, L36.

42 C H A P T E R 1

[57] X. Liu, Z. Xu, Y. Liu, Y. Shen, A novel high performance ethanol gas sensor based on CdO-Fe2O3 semiconducting materials, Sens. Actuators B, 1998, 52, 270-273.

[58] V. Guidi, M. C. Carotta, M. Ferroni, G. Martinelli, L. Paglialonga, E. Comini, G. Sberveglieri, Preparation of nanosized titania thick and thin films as gas sensors, Sens. Actuators B, 1999, 57, 197-200.

[59] M. Ferroni, D. Boscarino, E. Comini, D. Gnani, V. Guidi, G. Martinelli, P. Nelli, V. Rigato, G. Sberverglieri, Nanosized thin films of tungsten-titanium mixed oxides as gas sensors, Sens. Actuators B, 1999, 58, 01. Mrz, 289-294.

[60] A. A. Tomchenko, Structure and Gas Sensing Properties of WO3-Bi2O3 Mixed Thick-Films, Eurosensors XIII Conference Proceedings, 1999, The Hague, Nether-lands, 69.

[61] L. Satyanarayana, C. V. G. Reddy, S. V. Manorama, V. J. Rao, Liquid-petroleum-gas sensor based on a spinel semiconductor, ZnGaO4, Sens. Actuators B, 1998, 46, 01. Jul.

[62] J. S. Han, T. Bredow, D. E. Davey, A. B. Yu, D. E. Mulcahy, The effect of Al addi-tion on the gas sensing properties of Fe2O3 based sensors, Sens. Actuators B, 2001, 75, 18-23.

[63] M. Fleischer, H. Meixner, Selectivity in high-temperature operated semiconductor gas-sensors, Sens. Actuators B, 1998, 52, 179-187.

[64] N. Barsan, A. Tomescu, The temperature dependence of the response of SnO2-based gas sensing layers to O2, CH and CO, Sens. Actuators B, 1995, 26, 3, 45-48.

[65] T. Takada, A new method for gas identification using a single semiconductor sen-sor, Sens. Actuators B, 1998, 52, 45-52.

[66] G. Tournier, C. Pijolat, R. Lalauze, B. Patissier, Selective detection of CO and CH4 with gas sensors using SnO2 doped with palladium Pd, Sens. Actuators B, 1995, 26, 1-3, 24-28.

[67] M. Jaegle, J. Wöllenstein, T. Meisinger, H. Böttner, G. Müller, T. Becker, C. Bosch-v.Braunmühl, Micromachined thin film SnO2 gas sensors in temperature-pulsed operation mode, Sens. Actuators B, 1999, 57, 130-134.

[68] A. Galdikas, A. Mironas, A. Setkus, Copper-doping level effect on sensitivity and selectivity of tin oxide thin-film gas sensor, Sens. Actuators B, 1995, 26, 1-3, 29-32.

[69] A. Galdikas, A. Mironas, D. Senuliene, A. Setkus, W. Göpel, K. D. Schierbaum, Copper on-top sputtering induced modification of tin dioxide thin film gas sen-sors, Sens. Actuators B, 1999, 58, 1-3, 330-337.

[70] G. Behr, W. Fliegel, Electrical properties and improvement of the gas sensitivity in multiple-doped SnO2, Sens. Actuators B, 1995, 26, 3, 33-37.

[71] E. Comini, G. Faglia, G. Sberveglieri, CO and NO2 response of tin oxide silicon doped thin films, Sens. Actuators B, 2001, 76, 1-3, 270-274.

[72] W.-Y. Chung, D.-D. Lee, B.-K. Sohn, Effects of added TiO2 on the characteristics of SnO2 based thick-film gas sensors, Thin Solid Films, 1992, 221, 304-310.

[73] C. D. Feng, Y. Shimizu, M. Egashira, Effect of Gas Diffusion Process on Sensing Properties of SnO2 Thin Film Sensors in a SiO2/SnO2 Layer-Built Structure Fabri-cated by Sol-Gel Process, J. Electrochem. Soc., 1994, 141, 221.

[74] A. D'Amico, C. D. Natale, Chemical Portraits of Gasses and Liquids: Trends and Perspectives, Eurosensors XIII Conference Proceedings, 1999, The Hague, Netherlands, 999.

I N T R O D U C T I O N 43

[75] H. Baltes, D. Lange, A. Knoll, The electronic nose in Liliput, IEEE Journal, 1998, 9, 35-38.

[76] H. T. Nagle, R. Gutierrez, S. S. Schiffman, The HOW and WHY of electronic noses, IEEE Spectrum, 1998, 9, 22-34.

[77] M. Faia, M. A. Pereira, A. M. Nunes, C. S. Furtado, Electronic Noses, a different Approach to the Sensitivity and Selectivity Issues, J. Europ. Ceram. Soc., 1999, 19, 883-886.

[78] J. V. Hatfield, A. R. Daniels, D. Snowden, K. C. Persaud, P. A. Payne, Develop-ment of a Hand Held Electronic Nose (H2EN), Eurosensors XIII Conference Pro-ceedings, 1999, The Hague, Netherlands, 215.

[79] S. M. Reddy, P. A. Payne, Effect of unmodified and derivatised poly(vinyl chloride) overlayers on the response of an electronic nose based on conducting polymers, Sens. Actuators B, 1999, 58, 536-543.

[80] F. Winquist, I. Lundström, P. Wide, The combination of an electronic tongue and an electronic nose, Sens. Actuators B, 1999, 58, 512-517.

[81] C. Delpha, M. Siadat, M. Lumbreras, An electronic nose using time reduced mod-elling parameters for a reliable discrimination of Forane 134a, Sens. Actuators B, 2001, 77, 1-2, 517-524.

[82] E. Llobet, J. Rubio, X. Vilanova, J. Brezmes, X. Correig, J. W. Gardner, E. L. Hines, Electronic nose simulation tool centred on PSpice, Sens. Actuators B, 2001, 76, 1-3, 419-429.

[83] M. A. Martin, J. P. Santos, J. A. Agapito, Application of artificial neural networks to calculate the partial gas concentrations in a mixture, Sens. Actuators B, 2001, 77, 1-2, 468-471.

[84] D. L. Massart, B. G. M. Vandeginste, "Chemometrics: a textbook", Elsevier, Amsterdam, 1988.

[85] T. Eklov, I. Lundstrom, Gas mixture analysis using a distributed chemical sensor system, Sens. Actuators B, 1999, 57, 274-282.

[86] H. K. Hong, H. W. Shin, H. S. Park, D. H. Yun, C. H. Kwon, K. Lee, S. T. Kim, T. Moriizumi, Gas identification using oxide semiconductor gas sensor array and neural-network pattern recognition, Transducers 95, Eurosensors IX, 1995, 687.

[87] B. S. Joo, N. J. Choi, Y. S. Lee, J. W. Lim, B. H. Kang, D. D. Lee, Pattern recogni-tion of gas sensor array using characteristics of impedance, Sens. Actuators B, 2001, 77, 1-2, 209-214.

[88] D. S. Lee, J. K. Jung, J. W. Lim, J. S. Huh, D. D. Lee, Recognition of volatile organic compounds using SnO2 sensor array and pattern recognition analysis, Sens. Actuators B, 2001, 77, 1-2, 228-236.

[89] G. Reinhardt, S. I. Somov, U. Schönauer, U. Guth, W. Göpel, Solid electrolytes for gas sensing at high temperatures. Multielectrode setup to analyze gas mixtures, Transducers, Eurosensors IX, 1995, 799.

[90] P. G. Harrison, N. C. Lloyd, W. Daniell, C. Bailey, W. Azalee, Evolution of Micro-structure during the Thermal Activation of Chromium-Promoted Tin(IV) Oxide Catalysts: An FT-IR, FT-Raman, XRD, TEM and XANES/EXAFS Study, Chem. Mater., 1999, 11, 896-909.

44 C H A P T E R 1

[91] N. Barsan, M. Schweizer-Berberich, W. Göpel, Fundamental and practical aspects in the design of nanoscaled SnO2 gas sensors: a status report, Fresenius J. Anal. Chem., 1999, 365, 287-304.

[92] N. Barsan, U. Weimar, Conduction Model of Metal Oxide Gas Sensors, J. Electro-ceram., 2001, 7, 143-167.

[93] N. Barsan, A. Heilig, J. Kappler, U. Weimar, W. Göpel, CO-Water Interaction with Pd-doped SnO2 Gas Sensors: Simultaneous Monitoring of Resistances and Work Functions, Eurosensors XIII Conference Proceedings, 1999, The Hague, Nether-lands, 367.

[94] R. Ionescu, Ageing and p-type conduction in SnO2 gas sensors, Sens. Actuators B, 1999, 58, 1-3, 375-379.

[95] S. Nicoletti, L. Dori, F. Corticelli, M. Leoni, P. Scardi, Tin Oxide Thin-Film Sen-sors for Aromatic Hydrocarbons Detection: Effect of Aging Time on Film Micro-structure, J. Am. Ceram. Soc., 1999, 82, 1201.

[96] A. Dieguez, J. L. Alay, A. Cabot, A. Romano-Rodriguez, J. R. Morante, J. Kappler, N. Barsan, U. Weimar, W. Göpel, Influence on the gas sensor performance of the metal chemical states introduced by impregnation of SnO2 sol-gel nanocrystals, Eurosensors XIII Conference Proceedings, 1999, The Hague, Netherlands, 109.

[97] J. Kappler, N. Barsan, U. Weimar, A. Dieguez, J. L. Alay, A. Romano-Rodriguez, J. R. Morante, W. Göpel, Correlation between XPS, Raman and TEM measurements and the gas sensitivity of Pt and Pd doped SnO2 based gas sensors, Fresenius J Anal Chem, 1998, 361, 110-114.

[98] R. A. Bennett, P. Stone, M. Bowker, Pd nanoparticle enthanced re-oxidation of non-stoichiometric TiO2 STM imaging of spillover and a new form of SMSI, Catalysis Letters, 1999, 59, 99-105.

[99] P. Ingleby, J. W. Gardner, P. N. Bartlett, Effect of micro-electrode geometry on response of thin-film poly(pyrrole) and poly(aniline) chemoresistive sensors, Sens. Actuators B, 1999, 57, 17-27.

[100] S. B. Basame, H. S. White, Scanning Electrochemical Microscopy of Metal/Metal Oxide Electrodes. Analysis of Spatially Localized Electron-Transfer Reactions dur-ing Oxide Growth, Anal. Chem., 1999, 71, 3166-3170.

[101] A. Chiorino, G. Ghiotti, F. Prinetto, M. C. Carotta, D. Gnani, G. Martinelli, Preparation and characterization of SnO2 and MoOx-SnO2 nanosized powders for thick-film gas sensors, Sens. Actuators B, 1999, 58, 3, 338-349.

[102] M. C. Carotta, M. Ferroni, D. Gnani, V. Guidi, M. Merli, G. Martinelli, M. C. Casale, M. Notaro, Nanostructured pure and Nb-doped TiO2 as thick film gas sensors for environment gas monitoring, Sens. Actuators B, 1999, 58, 3, 310-317.

[103] G. Martinelli, M. C. Carotta, E. Traversa, G. Ghiotti, Thick-Film Gas Sensors Based on Nano-Sized Semiconducting Oxide Powders, MRS Bulletin, 1999, 24, 6, 30-36.

[104] G.-J. Li, S. Kawi, High Surface SnO2: a novel semiconductor-oxide gas sensor, Mat. Lett., 1998, 34, 99-102.

[105] J. S. Suehle, R. E. Cavicchi, M. Gaitan, S. Semancik, Tin Oxide Gas Sensor Fabri-cated Using CMOS Micro-Hotplates and In-Situ Processing, IEEE Electron Device Letters, 1993, 14, 3, 118-120.

[106] C. Hagleitner, A. Hierlemann, D. Lange, A. Kummer, N. Kerness, O. Brand, H. Baltes, Smart single-chip gas sensor system, Nature, 2001, 414, 293-296.

I N T R O D U C T I O N 45

[107] D. Briand, A. Krauss, B. van der Schoot, U. Weimar, N. Barsan, W. Gopel, N. F. de Rooij, Design and fabrication of high-temperature micro-hotplates for drop-coated gas sensors, Sens. Actuators B, 2000, 68, 1-3, 223-233.

[108] F. Solzbacher, C. Imawan, H. Steffes, E. Obermeier, H. Möller, A modular system of SiC-based microhotplates for the application in metal oxide gas sensors., Sens. Actuators B, 2000, 64, 95-101.

[109] D. Briand, B. v. d. Schoot, N. F. d. Rooij, H. Sundgren, I. Lundström, A Low-Power Micromachined MOSFET Gas Sensor, IEEE J. MEMS, 2000, 9, 3, 303-308.

[110] A. Heilig, N. Barsan, U. Weimar, W. Göpel, Selectivity enhancement of SnO2 gas sensors: simultaneous monitoring of resistance and temperatures, Sens. Actuators B, 1999, 58, 1-3, 302-309.

[111] V. Guidi, G. C. Cardinali, L. Dori, G. Faglia, M. Ferroni, G. Martinelli, P. Nelli, G. Sberveglieri, Thin-film gas sensor implemented on a low-power-consumption micromachined silicon structure, Sens. Actuators B, 1998, 49, 88-92.

[112] M. C. Horrillo, I. Sayago, L. Ares, J. Rodrigo, J. Gutierrez, A. Götz, I. Garcia, L. Fonseca, C. Cane, E. Lora-Tamayo, Detection of low NO2 concentrations with low power micromachined tin oxide gas sensors, Sens. Actuators B, 1999, 58, 1-3, 325-329.

[113] Z. Tang, S. K. H. Fung, D. T. W. Wong, P. C. H. Chan, J. K. O. Sin, P. W. Che-ung, An integrated gas sensor based on tin oxide thin-film and improved micro-hotplate, Sens. Actuators B, 1998, 46, 174-179.

[114] D. Vincenzi, M. A. Butturi, M. Stefancich, C. Malagu, V. Guidi, M. C. Carotta, G. Martinelli, V. Guarnieri, S. Brida, B. Margesin, F. Giacomozzi, M. Zen, A. Vasiliev, A. V. Pisliakov, Low-power thick-film gas sensor obtained by a combina-tion of screen printing and micromachining techniques, Thin Solid Films, 2001, 391, 288-292.

[115] K. Steiner, G. Sulz, E. Neske, E. Wagner, Laser annealing of SnO2 thin-film gas sensors in single chip packages, Sens. Actuators B, 1995, 26 - 27, 64-67.

46 C H A P T E R 1

47

M. Heule, S. Vuillemin, L. J. Gauckler,submitted as review article.

Overview on Powder-based Ceramic Meso- and Microscale Fabrication Processes

Processing techniques are reviewed that allow the introduction of ceramic components made from powders into microelectromechanical systems (MEMS). Ceramics have several advantages over other materials also in microsystems, e.g. heat resistance, hardness, corrosion resistivity or func-tional properties. The range of available materials in microfabrication technology is being increased beyond those deposited by thin-film technol-ogy. Top-down approaches like mechanical and laser-based direct writing processes, ink-jet printing, microextrusion and lithography-based methods are presented. They are complemented by some more fundamental work in the field of bottom-up synthesis of micro- and nanoscaled ceramic materi-als.

2.1. Introduction

Ceramic materials deposited as thin films have been used extensively in elec-tronics and MEMS (micro-electromechanical systems). However, ceramicpowders have not been considered for microfabrication processes untilrecently. Ceramics have several advantages over other materials in MEMS:[1]

heat resistance, hardness, corrosion resistivity even in harsh environments,chemical inertness for biological applications and abundant functional prop-erties like piezoelectricity,[2,3] pyroelectricity or catalytic activity of surfaces.Many MEMS devices could be improved if ceramic thick-film coatings asmicrostructures were readily available, e.g. Harris et al. modeled and fabri-

2.

48 C H A P T E R 2

cated cross flow micro heat exchangers. The need for an analogous LIGAstructure made of ceramics to the present PMMA polymer structure wasclearly stated. Better heat exchange properties and durability of ceramicscould be expected.[4] In another project concerning the fabrication ofMEMS-based microthrusters for application in small space crafts, ceramicswould be much better thermal insulators and enhance lifetime.[5,6] Cur-rently, the microthrusters are fabricated entirely in silicon.

The aim of this work is to review recently introduced microfabricationprocesses for small ceramic objects in the range from 1 mm down to the µmsize range. Most methods presented rely on a top-down approach, i.e. theyare based on scaling down an existing larger-scale fabrication method. A sum-mary of the presented processes is given in Table 2–1.

Table 2–1. Overview on microfabrication processes with ceramic powders. The list is ordered by the lowest achievable resolution that was reported or demonstrated in the referenced articles.

Method

Smallest Feature Resolution /µm

Aspect ratio achieved

2D-layers or 3D-bodiesa

References

STM tip electrochemical etching

0.01 - 2D [9]

Casting suspensions into standard photolithogra-phy masks

1-5 1-2 2D [72], [73]

Soft Lithography 1-5 1-3 2D/3D [74], [75], [76], [77], [80], [81]

Microstereolithography (UV-curable polymer solu-tion filled with alumina)

2 high 3D [42]

Co-extrusion (µm resolu-tion in two dimensions, rods mm long)

5-16 high 3D [49], [50]

MAPLE direct write 10 low 2D [36], [37], [38]

LIGA 10-20 10 2D [65]

LTTC-ML (low-temperature co-fired ceramic multi-layer) technology

25-100 variable (stacking)

3D [61], [62], [64]

M I C R O S C A L E C E R A M I C P R O C E S S E S 49

Direct Ceramic Machining DCM of presintered bodies

50 variable 3D [47]

Screen Printing 100 low 2D [44]

Pulsed Laser Ablation (depending on the optics used)

30-200 high 2D [29], [30]

Precision grinding (micro-cylindrical shapes)

50 high 3D [57]

Metal embossing 50 high 2D [52]

Ink-jet printing of suspen-sions

70 low 2D [14]

Freeform Ink-jet printing of suspensions (smallest wall thickness)

170 high 3D [13]

3DPTM process (ink-jet printing of binder solution into dried powder)

200 variable 3D [15], [16]

Micropen Writing (free-standing 3D Periodic Struc-tures)

250 1 3D [21], [25], [26], [27]

a. 2D-layers: substrate-supported structures, 3D-bodies: possibility to create freestanding microparts.

Table 2–1. Overview on microfabrication processes with ceramic powders. The list is ordered by the lowest achievable resolution that was reported or demonstrated in the referenced articles.

Method

Smallest Feature Resolution /µm

Aspect ratio achieved

2D-layers or 3D-bodiesa

References

50 C H A P T E R 2

2.2. Direct writing methods

Direct writing systems consist of an automated translation stage which movesa pattern generating device like an ink-jet head or laser writing optics. A gen-eral review on the state of the art in ink-jet printing of different materials canbe found elsewhere.[7] Structures are drawn in sequence and formed continu-ously. In contrast to parallel processes like photolithography, each sample hasto be fabricated individually which puts certain limits on the fabricationspeed and efficiency. On the other hand, each device can be built with anindividual shape with computer-controlled movement. Certain approachesallow the fabrication of complex shapes in all three dimensions which is oftentermed freeform fabrication.[8] As an example of ultimately small direct writ-ing, Hung et al. used an STM tip to electrochemically etch a square-shapedhole of exactly 1 µm side length into Tl2O3 thin films.[9] When a bias volt-age of -2.5 V is applied between tip and surface, etching of Tl2O3 is per-formed. Within one minute, a 200 nm grain under the tip could bedissolved.

2.2.1. Direct writing of powders using ink jet systems

Ainsley et al. reported a method for solid freeform fabrication (the manufac-turing of a small part in 3D) by controlled droplet deposition of powderfilled melts. Al2O3 powders were suspended in n-alkane mixtures of meltingtemperatures from 50-60°C. Deposited by an ink-jet system, free-standingbodies like a rotation wheel of 1-2 cm,[10,11] or test structures with a mini-mal wall thickness of 100 µm in the green state were demonstrated.[12] Thesolids loading could be increased up to 40 vol%. Zhao et al. similarly usedZrO2 powders suspended in organic matrices for the generation of mazestructures with vertical side walls within 170 ± 10 µm in the sintered state(Fig. 2–1a).[13]

The same group was also able to ink-jet print aqeous lead zircon titanate(PZT) suspensions with a particle size of 37.5 µm in two dimensions ontofilter paper. There, the resolution of 360 dpi was given by the use of a com-mercial ink-jet printer. Other limitations were found in the layer thickness,samples thicker than 20 layers could not be dried in good quality.[14]

M I C R O S C A L E C E R A M I C P R O C E S S E S 51

Alternatively, only binder solution without particles can be ink-jet printed toconsolidate a well defined region in a dried layer of powder.[15-17] The nextlayer is then sprayed as a suspension and left to dry before binder solution isprinted again for the definition of the part shape. The process is called3DPTM (Three Dimensional Printing) and has been developed at MIT.[18]

Complex-shaped 3D parts similar to those from stereolithography can bemanufactured using 3DP. The exact conditions of binder solution soaking aporous powder bed had to be characterised thoroughly.[19] Typical dimen-sions of such a disk-shaped volume element formed by one droplet of bindersolution lie in the range of 50 µm for the thickness and 200 µm for the diskdiameter. This volume element is also the smallest feature size that can bereproduced by the process. For removal of the powder regions not consoli-dated by binder solution, the part is immersed in a water bath.[20]

Fig. 2–1. a) ink jet printing of ZrO2 structures with a 170 µm resolution. The inks have a solid loading of 14 vol%.[13] b) MAPLE-direct-write process (matrix-assisted pulsed-laser evaporation). A variety of oxides, metals for electrical contacts and polymers can be trans-ferred.[36] c) threedimensionally layered mesostructures of silica colloids fabricated by dip-pen writing.[21]

substrate

ribbon

hνcoating material objective

a) b) MAPLE direct writing

c) dip-pen writing

substrate

250 µm

ink-jet printing of colloids

170 µm

substrate

oil immersion

ZrO2

SiO2

ink-jet head

52 C H A P T E R 2

2.2.2. Micropen writing

Another direct writing technique is called micropen fabrication in which apowder-filled paste is extruded through a small orifice. Smay et al. demon-strated the power of the approach by writing 3D-periodic structures whichconsist of layers of gratings stacked perpendicularly on each other (Fig. 2–1c).[21] This requires that the deposited structures of 250 µm width have tosupport their own weight as deposited for bridging the gaps in the underlyinglayer(s). Therefore, specifically tailored colloidal systems[22-24] to satisfy thatneed were developed by using e.g. poly-ethylenimine coated silica micro-spheres as a viscoelastic gel obtained by setting the pH to the isoelectric pointat pH 10. In order to slow the drying, the deposition was carried outimmersed in a non-wetting oil. In a similar fashion, these results were trans-ferred to aqeous PZT suspensions under different pH conditions.[25] Forassessment of the maximum spanning distance, angled test structures withincreasing width were deposited. The strongest colloidal gel at pH 6.15 wasable to span up to 2 mm distance with a buckling of less than 500 µm. Byusing gel-casting (stabilisation is reached by in-situ polymerisation of thepaste after deposition), freestanding structures of 250 µm thickness spanningseveral mm have been micropen-deposited.[26] As first applications,Pb(Nb,Zr,Ti)O3 pastes were deposited as thick-film capacitors,[27] and ZnOvaristors written by this method were characterised compared to standardZnO varistors.[28]

2.2.3. Laser-based direct writing

Using pulsed laser systems, enough energy can be focused on a spot to ablateeven the hardest materials. Kruusing et al. used Nd:YAG lasers to cut outmagnetic ceramics NdFeB and MnZn ferrites.[29] Cut sizes ranged from 30-50 µm. Oliveira et al. observed the formation of very small columns whileablating Al2O3 ceramics with a KrF excimer laser.[30] The typical craterformed by one laser pulse is of cylindric shape with a diameter of approx.200 µm and a depth of 30 µm. They also characterised the plasma plumesformed of the ablated ceramic material using optical emission spectros-copy.[31] Laser ablation may leave a different chemical composition of thesurface behind as Laude et al. have shown upon ablating Y-Si-Al-O-N ceram-

M I C R O S C A L E C E R A M I C P R O C E S S E S 53

ics.[32] They observed melting, Si3N4 decomposition and enrichment ofSiO2 on the surface. On the application level, Hanreich et al. fabricatedultra-thin pick-up coils for magnetic flux detection by filling laser-drilledholes in an Al2O3 substrate with electrode material.[33] This drilling made itpossible to create flat alumina-integrated electrodes which can be put intoclose contact with a magnetic surface to measure its flux. The holes are fun-nel-shaped with an outer diameter of 700 µm and an inner diameter of210 µm. Furthermore, this drilling method was used for creating human skinhumidity sensors requiring a flat, integrated electrode design to contactskin.[34]

Since many ceramic materials are chemically inert, etching is a difficultapproach. However, Horisawa et al. enhanced the etching of Al2O3 ceramicsimmersed in H3PO4 locally by a pulsed laser and were able to create smallholes of 0.5 mm diameter.[35]

A very elegant method to transfer ceramic materials from one substrate toanother has been developed by the Naval Research Laboratories in Washing-ton DC. It was termed MAPLE direct write technique (matrix assisted pulsedlaser evaporation).[36-38] The material to be deposited is coated on a UV-transparent sheet of quartz serving as a ribbon. The ribbon is then broughtinto close contact with an arbitrary, flat substrate of metal, plastics or ceram-ics (Fig. 2–1b). With a pulsed UV-laser, the interface between ribbon andcoating is quickly heated, causing a rapid ablation of matrix material whichin turn propels the coating material towards the substrate. If ceramic powdersare to be deposited, the powder is embedded in an organic binder matrix.This has been demonstrated using BaTiO3 and Au particles on two separateribbons. The gold was used to write electrodes, subsequently, a small capaci-tor structure of 15 µm thickness was deposited with the barium titanate. Therange of materials available is impressive,[39] also polymers and even biomole-cules and eukaryotic cells were transferred.[40] These complex molecules andcells are not harmed since they are embedded in a polymer matrix whichabsorbs almost all laser power and is pyrolysed. The expanding plume ofgases carries the complex molecules into the gas phase. The same mechanismis used for soft ionisation in MALDI mass spectrometry (Matrix AssistedLaser Desorption and Ionisation). Materials limitations of the method arise if

54 C H A P T E R 2

complex oxides are to be transferred. They may partially be decomposedupon ablation. Additionally, the same laser was used to perform laser-inducedannealing or trimming of the deposited layers. Fitz-Gerald et al. demon-strated the possibility to deposit phosphor powders consisting of Y2O3:Eu orZn2SiO4:Mn for high-definition display applications.[41]

In stereolithography, a laser is scanned across a UV-curable polymer solu-tion. Complex-shaped 3D parts can be manufactured if the hardened part islowered, exposing fresh solution into which the next layer is written. Suchpolymer solutions were filled by Zhang et al. with ceramic powders, e.g.33 vol% Al2O3 with a diameter of 200 nm.[42] The curing reaction thenforms a green body. The lateral resolution was improved down to 2 µm usingappropriate focussing optics. 15 µm thick single layers on a substrate couldbe obtained. Full ceramic microgears with diameters of 400 to 1000 µm(20 µm thick) were generated and could be sintered at 1400°C, exhibiting alow shrinkage.

Provin et al. used a similar scheme to fabricate very smooth 3D-shapedparts in the mm range.[43] In contrast to the previously cited work, UV-lightfrom a Hg lamp is projected via a LCD screen with VGA resolution(640x480 pixels) which is used to define the pattern of a single layer in oneexposure step. Solids loading of alumina was 24 vol%. It was observed thatthe UV-absorption increases due to the alumina filling which limits the cur-ing depth. Consequently, a depth resolution of < 10 µm was reached.

2.3. Downsizing Mechanical Processing Methods

Silk screen printing is a process of choice for many thick-film coating appli-cations since it is available up to industrial scale. Typical film thicknessesrange from 5 to 50 µm. The main drawback is the limited lateral resolutionaround 100 µm which is a function of mesh size and ceramic paste proper-ties.[44] Thiele et al. examined process compatibility issues between screen-printed PZT layers and silicon substrates.[45] Traditional machining of sintered ceramics is done with diamond-cuttingtools. Conventional tools can be used when using a soft presintered ceramicbody. If the presintered blank is homogeneous, it shrinks linearly in alldimensions and makes it possible to precompensate the sintering shrinkage

M I C R O S C A L E C E R A M I C P R O C E S S E S 55

by machining enlarged features. A final pattern reproduction quality of 0.1%length deviation in 50 µm structures can be obtained by this Direct CeramicMachining (DCM) process.[46,47]

The application of piezoelectric ceramics as ultrasonic transducers hasbeen the driving force for the development of various methods to create poly-mer-filled arrays of small ceramic rods with a high aspect ratio.[48] It wasshown that the resolution in medical ultrasonic imaging and related technol-ogies can be improved by miniaturizing the rod size. Among the techniquesare rod placing, dice and fill, lost mold processes (mechanically engineeredmolds), injection molding, laser cutting, solvent-air jet machining of greentapes and ceramic fiber processing. The smallest rods were 20 - 50 µm.

The group of Halloran has developed a multistage co-extrusion processwhich was used to extrude alumina, PZT and Pb-Mg-Nb-TiO3 pastes(Fig. 2–2a).[49] Embedded in a carbon-black matrix, M-shaped ceramicextrudates with a cross section of 2 mm were stacked 5 by 5 to be extrudedagain. Together, these 25 M-shapes were thereby reduced in size 5:1. If thisprocess of reassembly and extrusion is repeated, the size of the M-shape isreduced by a factor of five each time n, whereas the number of M-shapesincreases by 25n. After four reduction steps, an extrudate containing 390,625M-shapes of 16 µm height were obtained. Using the same size reduction pro-cess, other ceramics-metal composites with a specific square micro-configura-tion in the order of 15 µm for the ceramic phase and 5 µm for the AgPdmetallic interlayers were produced.[50]

On a larger scale, Knitter et al. presented a rapid prototyping for ceramicmicroreactors by low-pressure injection molding.[51] The molds with smallestfeatures of 500 µm were fabricated using stereolithography and moldingthem in silicone. In another work, they present PZT microrod arrays (50 µmwide, 300 µm high) for ultrasonic transducers using a combination of tapecasting and metal mold embossing.[52] The group of Lange used a process ofcolloidal isopressing against the 130 µm features of a coin which were repro-duced as an accurate surface replica in Al2O3 ceramics.[53,54] Similar surfacereproduction processes include the fabrication of titania scaffolds for hepato-cyte cell cultures[55] and the micro-embossing with a polymeric stamp intoshrinkage-free ZrSiO4 ceramics.[56] Alternatively, Yeo et al. produced micro-

56 C H A P T E R 2

cylindrical parts of 50 µm diameter with a precision grinding process.[57] Areactive hot isostatic pressing route using powders was adapted to manufac-ture SiC microrotors[58] and again PZT microstructures for ultrasonic trans-ducers.[59] Molds were fabricated by standard micromachining of a siliconwafer.

2.3.1. Meso-scaled devices based on co-fired ceramic tape technology

An important example of advanced ceramic technology suited for manufac-turing miniature devices is the LTTC process (Low Temperature Co-firedCeramic).[60] It is based on ceramic green tapes (“GT”) of 100-400 µmthickness which are soft and can be readily cut, ground or dissolved. They arecommercially available by Dupont Electronic Materials. Holes and groovesfor a microsystem are cut into such tape layers and by stacking them, com-plex structures can be built. Alternatively, smaller holes can be etched by dis-solving a photolithographically defined pattern. Electrical connections areintegrated by screen-printing on tape surfaces and by filling holes with metalpaste in the vertical direction. For final hardening, the green tape layers of thestack are first laminated by applying pressures up to 3600 psi at 60-80°C,then sintered at 875°C. In a recent overview by Gongora-Rubio, severalaspects of versatility and device applications are presented.[61] Small holes of25 µm can be “etched” into GT tapes by using an acetone containing jet ofnitrogen out of a small nozzle. Whereas binder organics are dissolved by theacetone, the force of the jet drives the filling powders away. One problemthat has to be addressed is the sagging of tape in the sealing process of emptycavities during the lamination process. Among the devices built were proxim-ity sensors, gas flow sensors, electrochemical sensors and pressure sensors.Vojak et al. created a ceramic microdischarge device with LTTC technology,housing a stable Ne-discharge in a cylindrical cavity of 140 µm diameter and230 µm length at a voltage of 137 V.[62] For organic synthesis, a ceramicmicroreactor for the methoxylation of methyl-2-furoate has been demon-strated.[63] Finally, Wilcox et al. demonstrated that the technology is matureby integrating a PEM (polymer electrolyte membrane) fuel cell designincluding methanol gas reforming.[64] They also entered the field of micro-fluidics,

M I C R O S C A L E C E R A M I C P R O C E S S E S 57

Fig. 2–2. a) Alumina M-shape fabricated by co-extrusion in a carbon-black matrix in the final sintered state.[49] b) Alumina micro-rods pre-pared by slip pressing of a LIGA metal mold into powder.[67]

producing channel, valve and pumping systems, that either worked magneto-hydrodynamically or by piezo actuation. A continuous flow PCR (poly-merase chain reaction) device was also presented.

2.4. Lithography-based Processes

The LIGA technology (German acronym for Lithography, Galvanoformungand Abformtechnik) was the first to produce high-aspect ratio structures inmicrofabrication using x-rays for pattern exposure. There are instances,where LIGA-produced molds were filled with ceramic materials.[65] Toreproduce the shapes in a casting process accurately, shrinkage and cracking

M

M

M

M

M

M

M

M1:5 Extrusion

extruded M-Shaped rods of Al2O3 paste in carbon black matrix.

M

M

M

M

M

M

M

M

M

M

M

M

M

M

M

M

1:5 ExtrusionM

M

M

M

M

M

M

M

M

M

M

M

M

M

M

M

Stack andrepeat extrusionprocess.

a)

b) slip pressing with high aspect ratio LIGA mold.

20 µm

Al2O3 suspensiondemold,fire

Al2O3 ceramics

58 C H A P T E R 2

problems have to be solved. Chan et al. used a TEOS (tetraethoxysilane) solfilled with 50 nm titania particles for casting into PMMA-molds. The micro-gears obtained had single teeth of 50 µm.[66] Using the high-aspect ratio fea-tures of LIGA structures in a slip pressing scheme, Bauer et al. made Al2O3structures such as profile nozzles and posts in the order of 100 µm with abso-lutely vertical side walls and sharp edges (~ 1 µm, Fig. 2–2b).[67,68] Ruzzu etal. combined ceramic and metallic LIGA-structures for advanced electronicpackaging applications.[69] Feiertag et al. made use of an unique feature of x-ray lithography, the possibility to expose from different angles.[70] By thisscheme, angled and 3-dimensionally crossed structures for creating photoniccrystals were obtained. More and more, high-aspect ratio lithography basedon UV-curable resists like the epoxy-based SU-8 negative resists is beginningto replace LIGA.[71]

Standard UV-photoresists may also be used to cast ceramic structuresdirectly into photoresist masks.[72,73] Small thick-film structures of Al2O3 orSnO2 with an aspect ratio of 1 could readily be fabricated using a resist withhigher viscosity that forms layers of 5-15 µm in one spin-coating step.

A few years ago, Whitesides suggested the use of elastomeric silicones(mostly polydimethylsiloxane PDMS) for micropattern transfer, called softlithography.[74-77] There are several variations available and suited for directpatterning of liquids. Microcontact printing is used for stamping self-assem-bled monolayers serving as resist or as functional layers. In micromolding incapillaries (MIMIC), a fluid is patterned by spontaneous filling of PDMSmicrochannels. MIMIC was used to pattern polymeric precursor solutionsfor SiBNC ceramics[78] or Sr2Nb2O7 ceramics by a sol-gel route,[79] as wellas to structure suspensions of tin oxide,[80,81] using channel sizes down to10 µm. Such microlines of tin oxide were then integrated to form tiny semi-conducting gas sensors.[82,83] Yang et al. made pattern of porous oxides usingMIMIC.[84] By exploiting the flow characteristics in small channels, suchsystems were used to guide the growth of tubular silica structures.[85] PDMSmolds were also used for filling with ceramic suspensions, either generatingaccurately textured surfaces of bulk ceramics[86] or freestanding parts.[87]

M I C R O S C A L E C E R A M I C P R O C E S S E S 59

Using pre-ceramic polymers, a commercially available mixture (Ceraset) thatforms SiCN ceramics upon pyrolysis, the group of Raj exploited severalmicrofabrication schemes. Microgears were obtained by filling SU-8 pat-tern.[88] By adding pressure in a similar process called microforging, 20 µmwide lines could be shaped.[89] Moreover, Ceraset could be exposed directlythrough a photomask for photopolymerisation, if mixed with an appropriatephotoinitiator.[90] Feature sizes obtained are in the 100 µm range.

There are doped Li2O-Al2O3-SiO2 glasses that are photo-curable. Salimet al. successfully exposed and etched microgrippers a few mm long with fea-tures of approx. 300 µm.[91]

Due to their chemical inertness, ceramics are not an ideal material forphotolithography/etching processes. Makino et al. etched alumina, siliconnitride, SiC and others in phosphoric acid. For masking, non-standard UV-resists like polybutadiene rubbers were investigated. As a result, smoothtrenches of 100 µm to 600 µm and a depth of 30-40 µm could be etchedinto alumina in 30 min.[92] Wang et al. used a modern gas phase etchingprocess for machining PZT with a SF6-based deep reactive ion etching pro-cess.[93]

2.5. Self-Assembly

So far, all approaches presented rely on top-down strategies, i.e. they arebased on miniaturising existing techniques of larger scales. Alternatively, bot-tom-up approaches have also been successfully applied, in which the abilityto form the desired microstructure is inherent to the smaller building blocks,e.g. colloidal particles. The particles “find” their place in the structure by self-assembly without the need for intervention from the outside when mixedappropriately. Although that self-assembly processes are especially fascinating,they have not been used much for MEMS fabrication. Self-assembly still is afield of science rather than engineering. There might also be a lack of cross-disciplinary communication between MEMS development and colloidalchemistry research.

Photonic band gap materials are often prepared by self-assembly processesof colloidal crystals which are subsequently backfilled with a high refractiveindex material, e.g. sol-gel ceramics.[94] It is also important to note that for

60 C H A P T E R 2

application as useful photonic bandgap materials, a sufficiently large differ-ence in refractive index between particles and the surrounding phase in thecrystal is necessary. A property which again calls for the use of ceramic mate-rials.[95] Colloidal crystals consist of ordered colloidal particles in a cubicclosest package with a periodicity in the wavelength range of visible light(Fig. 2–3). Hayward et al. have further directed the assembly of colloidalcrystals by electrophoretic deposition into microstructures defined by theunderlying structured electrodes.[96,97] They also patterned BaTiO3 layersusing self-assembling block-copolymers that form nanoscale modulations ona surface.[98] In a recent review, Dabbs and Aksay show the abundant possi-bilities of shaping ceramic materials on a true nanometer scale. Parallels tobio-mineralisation and the hierarchical construction of biominerals aredrawn.[99] Vlasov et al. observed that a spherical shape of particles and a nar-row distribution of sphere size seems to be sufficient for spontaneously form-ing colloidal crystals.[100] In this study, silica spheres were used. In Xia’sgroup, a flow directed assembly of particles into colloidal crystals is pre-sented.[101-103]

Fig. 2–3. Silica particles in hexagonal close packing as templates for photo-nic band gap materials prepared by micromolding in capillaries.

M I C R O S C A L E C E R A M I C P R O C E S S E S 61

Out of all various approaches for the fabrication of photonic crystals, thestrikingly easy to reproduce convection self-assembly method suggested byColvin’s group seems to have become a standard method. The surface to becoated is placed into a dilute suspension of microspherical particles in etha-nol. As the solvent dries away under defined conditions, the particles assem-ble at the receding ethanol meniscus and form closest packed structures.[104-

109]

2.6. Summary

There are abundant approaches for generating small ceramic elements usingpowders as building blocks. Due to the scientific and technological experi-ence in synthesis and control of many functional properties, powders mayoffer enhanced functionalities also for MEMS devices compared to thinfilms. However, MEMS devices with successful integration of ceramic pow-ders in any form remain scarce. The LTTC green tape process is sophisticatedenough for making useful meso-scaled ceramic devices. For objects in therange of mm to 0.1 mm there are many processes available. Componentswith high aspect ratios of 100 or more are best molded using LIGA or relatedprocesses. Precision mechanical machining enables the fabrication of compo-nents with features of 50 to 100 µm with high aspect ratios at low cost. Con-siderably smaller structures in ceramics than 50 µm can be generated byMAPLE direct writing, soft lithography, standard photolithography andSTM tip etching. Photonic band gap materials with periodicities in the rangeof 100 to 1000 nm can be generated by a number of schemes. In most otherself-assembly approaches at the nanometer level, more development work isneeded for generating useful coatings and structures.

62 C H A P T E R 2

2.7. References

[1] L. A. Liew, W. G. Zhang, L. N. An, S. Shah, R. L. Luo, Y. P. Liu, T. Cross, M. L. Dunn, V. Bright, J. W. Daily, R. Raj, K. Anseth, Ceramic MEMS - New materials, innovative processing and future applications, Am. Ceram. Soc. Bull., 2001, 80, 5, 25-30.

[2] X. H. Zhu, J. M. Zhu, S. H. Zhou, Q. Li, Z. G. Liu, Microstructures of the mono-morph piezoelectric ceramic actuators with functional gradients, Sens. Actuator A-Phys., 1999, 74, 1-3, 198-202.

[3] Y. H. Lim, V. V. Varadan, V. K. Varadan, Finite-element modeling of the transient response of MEMS sensors, Smart Mater. Struct., 1997, 6, 1, 53-61.

[4] C. Harris, M. Despa, K. Kelly, Design and fabrication of a cross flow micro heat exchanger, J. Microelectromech. Syst., 2000, 9, 4, 502-508.

[5] C. Rossi, T. Do Conto, D. Esteve, B. Larangot, Design, fabrication and modelling of MEMS-based microthrusters for space application, Smart Mater. Struct., 2001, 10, 6, 1156-1162.

[6] C. Rossi, S. Orieux, B. Larangot, T. Do Conto, D. Esteve, Design, fabrication and modeling of solid propellant microrocket-application to micropropulsion, Sens. Actuator A-Phys., 2002, 99, 1-2, 125-133.

[7] P. Calvert, Inkjet printing for materials and devices, Chem. Mater., 2001, 13, 10, 3299-3305.

[8] J. W. Halloran, Freeform fabrication of ceramics, British Ceramic Transactions, 1999, 98, 6, 299-303.

[9] C. J. Hung, J. N. Gui, J. A. Switzer, Scanning probe nanolithography of conduct-ing metal oxides, Appl. Phys. Lett., 1997, 71, 12, 1637-1639.

[10] C. Ainsley, N. Reis, B. Derby, Freeform fabrication by controlled droplet deposi-tion of powder filled melts, J. Mater. Sci., 2002, 37, 15, 3155-3161.

[11] B. Derby, Materials opportunities in layered manufacturing technology, J. Mater. Sci., 2002, 37, 15, 3091-3092.

[12] K. A. M. Seerden, N. Reis, J. R. G. Evans, P. S. Grant, J. W. Halloran, B. Derby, Ink-jet printing of wax-based alumina suspensions, J. Am. Ceram. Soc., 2001, 84, 11, 2514-2520.

[13] X. Zhao, J. R. G. Evans, M. J. Edirisinghe, Direct Ink-Jet Printing of Vertical Walls, J. Am. Ceram. Soc., 2002, 85, 8, 2113-2115.

[14] J. Windle, B. Derby, Ink jet printing of PZT aqueous ceramic suspensions, J. Mater. Sci. Lett., 1999, 18, 2, 87-90.

[15] R. K. Holman, M. J. Cima, S. A. Uhland, E. Sachs, Spreading and infiltration of inkjet-printed polymer solution droplets on a porous substrate, J. Colloid Interface Sci., 2002, 249, 2, 432-440.

[16] S. A. Uhland, R. K. Holman, S. Morissette, M. J. Cima, E. M. Sachs, Strength of green ceramics with low binder content, J. Am. Ceram. Soc., 2001, 84, 12, 2809-2818.

[17] J. Moon, J. E. Grau, V. Knezevic, M. J. Cima, E. Sachs, Ink-Jet Printing of Binders for Ceramic Components, J. Am. Ceram. Soc., 2002, 85, 4, 755-62.

[18] J. Moon, J. E. Grau, V. Knezevic, M. J. Cima, E. M. Sachs, Ink-jet printing of binders for ceramic components, J. Am. Ceram. Soc., 2002, 85, 4, 755-762.

M I C R O S C A L E C E R A M I C P R O C E S S E S 63

[19] R. K. Holman, S. A. Uhland, M. J. Cima, E. Sachs, Surface adsorption effects in the inkjet printing of an aqueous polymer solution on a porous oxide ceramic sub-strate, J. Colloid Interface Sci., 2002, 247, 2, 266-274.

[20] J. Moon, J. E. Grau, M. J. Cima, E. M. Sachs, Slurry chemistry control to produce easily redispersible ceramic powder compacts, J. Am. Ceram. Soc., 2000, 83, 10, 2401-2408.

[21] J. E. Smay, G. M. Gratson, R. F. Shepherd, J. Cesarano, J. A. Lewis, Directed Col-loidal Assembly of 3D Periodic Structures, Adv. Mater., 2002, 14, 18, 1279-1283.

[22] J. A. Lewis, Colloidal processing of ceramics, J. Am. Ceram. Soc., 2000, 83, 10, 2341-2359.

[23] V. Tohver, A. Chan, O. Sakurada, J. A. Lewis, Nanoparticle engineering of com-plex fluid behavior, Langmuir, 2001, 17, 26, 8414-8421.

[24] J. A. Lewis, Binder removal from ceramics, Ann. Rev. Mater. Sc., 1997, 27, 147-173.

[25] J. E. Smay, J. Cesarano, J. A. Lewis, Colloidal inks for directed assembly of 3-D periodic structures, Langmuir, 2002, 18, 14, 5429-5437.

[26] S. L. Morissette, J. A. Lewis, J. Cesarano, D. B. Dimos, T. Y. Baer, Solid freeform fabrication of aqueous alumina-poly(vinyl alcohol) gelcasting suspensions, J. Am. Ceram. Soc., 2000, 83, 10, 2409-2416.

[27] S. L. Morissette, J. A. Lewis, P. G. Clem, J. Cesarano, D. B. Dimos, Direct-write fabrication of Pb(Nb,Zr,Ti)O(3 d)evices: Influence of paste rheology on print mor-phology and component properties, J. Am. Ceram. Soc., 2001, 84, 11, 2462-2468.

[28] V. Tohver, S. L. Morissette, J. A. Lewis, B. A. Tuttle, J. A. Voigt, D. B. Dimos, Direct-write fabrication of zinc oxide varistors, J. Am. Ceram. Soc., 2002, 85, 1, 123-128.

[29] A. Kruusing, S. Leppavuori, A. Uusimaki, B. Petretis, O. Makarova, Micromachin-ing of magnetic materials, Sens. Actuator A-Phys., 1999, 74, 1-3, 45-51.

[30] V. Oliveira, O. Conde, R. Vilar, UV laser micromachining of ceramic materials: Formation of columnar topographies, Adv. Eng. Mater., 2001, 3, 1-2, 75-81.

[31] V. Oliveira, J. C. Orlianges, A. Catherinot, O. Conde, R. Vilar, Laser ablation of Al2O3-TiC: a spectroscopic investigation, Appl. Surf. Sci., 2002, 186, 1-4, 309-314.

[32] L. D. Laude, C. Ogeret, A. Jadin, K. Kolev, Excimer laser ablation of Y-SiAlON, Appl. Surf. Sci., 1998, 129, 848-851.

[33] G. Hanreich, M. Mundleina, H. Hauser, J. Nicolics, G. Stangl, R. Grossinger, J. H. Espina-Hernandez, Ultra-thin pick-up coil for surface flux detection, Sens. Actuator A-Phys., 2001, 91, 1-2, 57-60.

[34] G. Hanreich, J. Nicolics, M. Mundlein, H. Hauser, R. Chabicovsky, A new bond-ing technique for human skin humidity sensors, Sens. Actuator A-Phys., 2001, 92, 1-3, 364-369.

[35] H. Horisawa, N. Akimoto, H. Ashizawa, N. Yasunaga, Effects of a quartz beam-guide on high-precision laser-assisted etching of Al2O3 ceramics, Surf. Coat. Tech-nol., 1999, 112, 1-3, 389-393.

[36] A. Pique, D. B. Chrisey, R. C. Y. Auyeung, J. Fitz-Gerald, H. D. Wu, R. A. McGill, S. Lakeou, P. K. Wu, V. Nguyen, M. Duignan, A novel laser transfer pro-

64 C H A P T E R 2

cess for direct writing of electronic and sensor materials, Applied Physics A, 1999, 69, S279-S284.

[37] D. B. Chrisey, A. Pique, R. Modi, H. D. Wu, R. C. Y. Auyeung, H. D. Young, Direct writing of conformal mesoscopic electronic devices by MAPLE DW, Appl. Surf. Sci., 2000, 168, 1-4, 345-352.

[38] D. B. Chrisey, A. Pique, J. Fitz-Gerald, R. C. Y. Auyeung, R. A. McGill, H. D. Wu, M. Duignan, New approach to laser direct writing active and passive mesos-copic circuit elements, Appl. Surf. Sci., 2000, 154, 593-600.

[39] A. Pique, D. B. Chrisey, J. M. Fitz-Gerald, R. A. McGill, R. C. Y. Auyeung, H. D. Wu, S. Lakeou, V. Nguyen, R. Chung, M. Duignan, Direct writing of electronic and sensor materials using a laser transfer technique, J. Mater. Res., 2000, 15, 9, 1872-1875.

[40] P. K. Wu, B. R. Ringeisen, J. Callahan, M. Brooks, D. M. Bubb, H. D. Wu, A. Pique, B. Spargo, R. A. McGill, D. B. Chrisey, The deposition, structure, pattern deposition, and activity of biomaterial thin-films by matrix-assisted pulsed-laser evaporation (MAPLE) and MAPLE direct write, Thin Solid Films, 2001, 398, 607-614.

[41] J. M. Fitz-Gerald, A. Pique, D. B. Chrisey, P. D. Rack, M. Zeleznik, R. C. Y. Auy-eung, S. Lakeou, Laser direct writing of phosphor screens for high-definition dis-plays, Appl. Phys. Lett., 2000, 76, 11, 1386-1388.

[42] X. Zhang, X. N. Jiang, C. Sun, Micro-stereolithography of polymeric and ceramic microstructures, Sens. Actuator A-Phys., 1999, 77, 2, 149-156.

[43] C. Provin, S. Monneret, Complex ceramic-polymer composite microparts made by microstereolithography, Ieee Transactions on Electronics Packaging Manufacturing, 2002, 25, 1, 59-63.

[44] M. J. Madou, "Fundamentals of Microfabrication - The Science of Miniaturisa-tion", 2nd ed., CRC Press, Boca Raton, 2002, 154.

[45] E. S. Thiele, D. Damjanovic, N. Setter, Processing and Properties of Screen-Printed Lead Zirconate Titanate Piezolelctric Thick Films on Electroded Silicon, J. Am. Chem. Soc., 2001, 84, 12, 2863-68.

[46] U. Schönholzer, F. T. Filser, P. Kocher, L. J. Gauckler, Fabrication of a Surface Pat-tern in Zirconia, Am. Ceram. Soc. Bull., 2000, 12, 45-47.

[47] L. J. Gauckler, P. Kocher, F. T. Filser, "Rapid Manufacturing of High-Tech Ceram-ics. A Case Study for Dental Applications", Shaping of Ceramics Conference Pro-ceedings, Gent (Be), Flemish Institute of Technological Research, 2002, 259-264.

[48] V. F. Janas, A. Safari, Overview of Fine Scale Piezoelectric Ceramic/Polymer Com-posite Processing, J. Am. Chem. Soc., 1995, 78, 11, 2945-55.

[49] C. Van Hoy, A. Barda, M. Griffith, J. W. Halloran, Microfabrication of ceramics by co-extrusion, J. Am. Ceram. Soc., 1998, 81, 1, 152-158.

[50] A. T. Crumm, J. W. Halloran, Fabrication of microconfigured multicomponent ceramics, J. Am. Ceram. Soc., 1998, 81, 4, 1053-1057.

[51] R. Knitter, D. Gohring, P. Risthaus, J. Hausselt, Microfabrication of ceramic microreactors, Microsyst. Technol., 2001, 7, 3, 85-90.

[52] R. Knitter, E. Günther, C. Odemer, U. Maciejewski, Ceramic microstructures and potential applications, Microsyst. Technol., 1996, 2, 135-138.

M I C R O S C A L E C E R A M I C P R O C E S S E S 65

[53] Z. Zhang, F. F. Lange, Patterning ceramic surfaces by colloidal isopressing, Adv. Eng. Mater., 2002, 4, 5, 294-295.

[54] B. C. Yu, F. F. Lange, Colloidal isopressing: A new shape-forming method, Adv. Eng. Mater., 2001, 13, 4, 276-280.

[55] S. Petronis, K. L. Eckert, J. Gold, E. Wintermantel, Microstructuring ceramic scaf-folds for hepatocyte cell culture, J. Mater. Sci.-Mater. Med., 2001, 12, 6, 523-528.

[56] V. D. Hennige, J. Hausselt, H. J. Ritzhaupt-Kleissl, T. Windmann, Shrinkage-free ZrSiO4-ceramics: Characterisation and applications, J. European Ceram. Soc., 1999, 19, 16, 2901-2908.

[57] S. H. Yeo, S. A. P. Balon, Fabrication of microcylindrical parts based on a novel grinding apparatus, Proc. Inst. Mech. Eng. Part B-J. Eng. Manuf., 2000, 214, 3, 245-249.

[58] J. F. Li, S. Sugimoto, S. Tanaka, M. Esashi, R. Watanabe, Manufacturing silicon carbide microrotors by reactive hot isostatic pressing within micromachined silicon molds, J. Am. Ceram. Soc., 2002, 85, 1, 261-263.

[59] S. A. Wang, J. F. Li, K. Wakabayashi, M. Esashi, R. Watanabe, Lost silicon mold process for PZT microstructures, Adv. Mater., 1999, 11, 10, 873.

[60] M. Gongora-Rubio, L. M. Sola-Laguna, P. J. Moffett, J. J. Santiago-Aviles, The utilization of low temperature co-fired ceramics (LTCC-ML) technology for meso-scale EMS, a simple thermistor based flow sensor, Sens. Actuator A-Phys., 1999, 73, 3, 215-221.

[61] M. Gongora-Rubio, P. Espinoza-Vallejos, L. Sola-Laguna, J. J. Santiago-Aviles, Overview of low temperature co-fired ceramics tape technology for meso-system technology (MsST), Sens. Actuator A-Phys., 2001, 89, 222-241.

[62] B. A. Vojak, S. J. Park, C. J. Wagner, J. G. Eden, R. Koripella, J. Burdon, F. Zen-hausern, D. L. Wilcox, Multistage, monolithic ceramic microdischarge device hav-ing an active length of similar to 0.27 mm, Appl. Phys. Lett., 2001, 78, 10, 1340-1342.

[63] V. Mengeaud, O. Bagel, R. Ferrigno, H. H. Girault, A. Haider, A ceramic electro-chemical microreactor for the methoxylation of methyl-2-furoate with direct mass spectrometry coupling, Lab Chip, 2002, 2, 39-44.

[64] D. L. Wilcox, J. Burdon, R. Changrani, C. Chou, S. Dai, R. Koripella, O. Manny, D. Sadler, P. von Allmen, F. Zenhausern In Materials Science of Microelectromechan-ical Systems (MEMS) Devices IV; A. A. Ayon, T. E. Buchheit, H. Kahn, S. M. Spear-ing, Eds.; Materials Research Society: Warrendale, PA, 2002; Vol. 687.

[65] W. Ehrfeld, V. Hessel, H. Lowe, C. Schulz, L. Weber, Materials of LIGA technol-ogy, Microsyst. Technol., 1999, 5, 3, 105-112.

[66] C. M. Chan, G. Z. Cao, T. G. Stoebe, Net shape ceramic microcomponents by modified sol-gel casting, Microsyst. Technol., 2000, 6, 5, 200-204.

[67] W. Bauer, H. J. Ritzhaupt-Kleissl, J. Hausselt, Micropatterning of ceramics by slip pressing, Ceramics International, 1999, 25, 3, 201-205.

[68] H. J. Ritzhaupt-Kleissl, W. Bauer, E. Günther, J. Laubersheimer, J. Hausselt, Development of Ceramic Microstructures, Microsyst. Technol., 1996, 2, 130-134.

[69] A. Ruzzu, O. Fromhein, D. Haller, Electrical bottom side contacting of electrome-chanical LIGA- structures by vias and screen printing, Microsyst. Technol., 2002, 8, 2-3, 113-115.

66 C H A P T E R 2

[70] G. Feiertag, W. Ehrfeld, H. Freimuth, H. Kolle, H. Lehr, M. Schmidt, M. M. Siga-las, C. M. Soukoulis, G. Kiriakidis, T. Pedersen, J. Kuhl, W. Koenig, Fabrication of photonic crystals by deep x-ray lithography, Appl. Phys. Lett., 1997, 71, 11, 1441-1443.

[71] H. Lorenz, M. Despont, N. Fahri, J. Brugger, P. Vettiger, P. Renaud, High aspect-ratio, ultrathick, negative-tone-near-UV-photoresist and its applications for MEMS, Sens. Actuator A-Phys., 1998, 64, 33-39.

[72] U. Schönholzer, R. Hummel, L. J. Gauckler, Microfabrication of Ceramics by Fill-ing of Photoresist Molds, Adv. Mater., 2000, 12, 17, 1261-1263.

[73] M. Heule, L. J. Gauckler, Microfabrication of Ceramics based on Colloidal Sus-pensions and Photoresist Masks, J. Photopolym. Sci. Technol., 2001, 14, 3, 449.

[74] Y. Xia, G. M. Whitesides, Soft Lithography, Angew. Chem. Int. Ed., 1998, 37, 550-575.

[75] E. Kim., Y. Xia, G. M. Whitesides, Polymer microstructures formed by moulding in capillaries, Nature, 1995, 376, 581-584.

[76] E. Kim, Y. Xia, G. M. Whitesides, Micromolding in Capillaries: Applications in Materials Science, J. Am. Chem. Soc., 1996, 118, 5722-5731.

[77] J. Hu, R. G. Beck, T. Deng, R. M. Westervelt, K. D. Maranowski, A. C. Gossard, G. M. Whitesides, Using soft lithography to fabricate GaAs/AlGaAs heterostruc-ture field effect transistors, Appl. Phys. Lett., 1997, 71, 14, 2020-2022.

[78] H. Yang, P. Deschatelets, S. T. Brittain, G. M. Whitesides, Fabrication of high per-formance ceramic microstructures from a polymeric precursor using soft lithogra-phy, Adv. Mater., 2001, 13, 1, 54.

[79] S. Seraji, Y. Wu, N. E. Jewell-Larson, M. J. Forbess, S. J. Limmer, T. P. Chou, G. Cao, Patterned Microstructure of Sol-Gel Derived Complex Oxides Using Soft Lithography, Adv. Mater., 2000, 12, 19, 1421.

[80] M. Heule, L. Meier, L. J. Gauckler, Micropatterning of Ceramics on Substrates towards Gas Sensing Applications, Mat. Res. Soc. Symp. Proc., 2001, Vol. 657, EE9.4.

[81] M. Heule, J. Schell, L. J. Gauckler, Powder-based Tin Oxide Micro-Components on Silicon Substrates Fabricated by Micromolding in Capillaries, J. Am. Ceram. Soc., 2002, in press.

[82] M. Heule, L. J. Gauckler, Gas Sensors Fabricated from Ceramic Suspensions by Micromolding in Capillaries, Adv. Mater., 2001, 13, 23, 1790.

[83] M. Heule, L. J. Gauckler, Miniaturised Arrays of Tin Oxide Gas Sensors on Single Microhotplate Substrates Fabricated by Micromolding in Capillaries, Sens. Actuator B, 2002, (in press).

[84] P. Yang, A. H. Rizvi, B. Messer, B. f. Chmelka, G. M. Whitesides, G. D. Stucky, Patterning Porous Oxides within Microchannel Networks, Adv. Mater., 2001, 13, No.6, 427.

[85] M. Trau, N. Yao, E. Kim, Y. Xia, G. M. Whitesides, I. A. Aksay, Microscopic pat-terning of orientated mesoscopic silica through guided growth, Nature, 1997, 390, 6661, 674-676.

[86] U. Schönholzer, L. J. Gauckler, Ceramic Parts Patterned in the Micrometer Range, Adv. Mater., 1999, 11, 8, 630-632.

M I C R O S C A L E C E R A M I C P R O C E S S E S 67

[87] B. Su, D. Zhang, T. W. Button, Micropatterning of fine scale ceramic structures, J. Mater. Sci., 2002, 37, 15, 3123-3126.

[88] L. A. Liew, W. G. Zhang, V. M. Bright, L. N. An, M. L. Dunn, R. Raj, Fabrication of SiCN ceramic MEMS using injectable polymer- precursor technique, Sens. Actu-ator A-Phys., 2001, 89, 1-2, 64-70.

[89] Y. P. Liu, L. A. Liew, R. L. Luo, L. N. An, M. L. Dunn, V. M. Bright, J. W. Daily, R. Raj, Application of microforging to SiCN MEMS fabrication, Sens. Actuator A-Phys., 2002, 95, 2-3, 143-151.

[90] L. A. Liew, Y. P. Liu, R. L. Luo, T. Cross, L. N. An, V. M. Bright, M. L. Dunn, J. W. Daily, R. Raj, Fabrication of SiCN MEMS by photopolymerization of pre-ceramic polymer, Sens. Actuator A-Phys., 2002, 95, 2-3, 120-134.

[91] R. Salim, H. Wurmus, A. Harnisch, D. Hulsenberg, Microgrippers created in microstructurable glass, Microsyst. Technol., 1997, 4, 1, 32-34.

[92] E. Makino, T. Shibata, Y. Yamada, Micromachining of fine ceramics by photoli-thography, Sens. Actuator A-Phys., 1999, 75, 3, 278-288.

[93] S. A. Wang, X. H. Li, K. Wakabayashi, M. Esashi, Deep reactive ion etching of lead zirconate titanate using sulfur hexafluoride gas, J. Am. Ceram. Soc., 1999, 82, 5, 1339-1341.

[94] J. D. Joannopoulos, R. D. Meade, J. N. Winn, "Photonic Crystals - Molding the Flow of Light", Princeton University Press, Princeton, NJ, 1995.

[95] O. D. Velev, E. W. Kaler, Structured Porous Materials via Colloidal Crystal Tem-plating: From Inorganic Oxides to Metals, Adv. Mater., 2000, 12, 7, 531.

[96] R. C. Hayward, D. A. Saville, I. A. Aksay, Electrophoretic assembly of colloidal crystals with optically tunable micropatterns, Nature, 2000, 404, 56-58.

[97] M. Trau, D. A. Saville, I. A. Aksay, Assembly of Colloidal Crystals at Electrode Interfaces, Langmuir, 1997, 13, 6375-6381.

[98] T. Lee, N. Yao, I. A. Aksay, Nanoscale Patterning of Barium Titanate on Block Copolymers, Langmuir, 1997, 13, 3866-3870.

[99] D. M. Dabbs, I. A. Aksay, Self-Assembled Ceramics Produced by Complex Fluid Templation, Annu. Rev. Phys. Chem., 2000, 51, 601-622.

[100] Y. A. Vlasov, X.-Z. Bo, J. C. Sturm, D. J. Norris, On-chip natural assembly of sili-con photonic bandgap crystals, Nature, 2001, 414, 289.

[101] S. H. Park, Y. Xia, Assembly of Mesoscale Particles over Large Areas and Its Appli-cation in Fabricating Tunable Optical Filters, Langmuir, 1999, 15, 266-273.

[102] S. H. Park, D. Qin, Y. Xia, Crystallization of Mesoscale Particles over Large Areas, Adv. Mater., 1998, 10, No. 13, 1028-1032.

[103] S. H. Park, Y. Xia, Macroporous Membranes with Highly Ordered and Three-Dimensionally Interconnected Spherical Pores, Adv. Mater., 1998, 10, No. 13, 1045-1048.

[104] J. F. Bertone, P. Jiang, K. S. Hwang, D. M. Mittleman, V. L. Colvin, Thickness dependence of the optical properties of ordered silica-air and air-polymer photonic crystals, Phys. Rev. Lett., 1999, 83, 2, 300-303.

[105] V. L. Colvin, From opals to optics: Colloidal photonic crystals, MRS Bull., 2001, 26, 8, 637-641.

68 C H A P T E R 2

[106] P. Jiang, G. N. Ostojic, R. Narat, D. M. Mittleman, V. L. Colvin, The fabrication and bandgap engineering of photonic multilayers, Adv. Mater., 2001, 13, 6, 389-393.

[107] P. Jiang, J. F. Bertone, V. L. Colvin, A lost-wax approach to monodisperse colloids and their crystals, Science, 2001, 291, 5503, 453-457.

[108] R. Rengarajan, P. Jiang, D. C. Larrabee, V. L. Colvin, D. M. Mittleman, Colloidal photonic superlattices, Phys. Rev. B, 2001, 6420, 20, art. no.-205103.

[109] M. E. Turner, T. J. Trentler, V. L. Colvin, Thin films of macroporous metal oxides, Adv. Mater., 2001, 13, 3, 180-183.

69

M. Heule, J. Schell, L. J. Gauckler,J. Amer. Ceram. Soc., 2003, in press.

Powder-Based Microcomponents on Silicon Substrates fabricated by Micromolding in Capillaries

With the introduction of soft lithography and micromolding in capillaries, low cost microfabrication with liquid materials has become possible. In this article, we demonstrate how to fabricate porous ceramic lines of 10 µm width and several mm in length on silicon wafer substrates by using colloidal suspensions of tin oxide. Microchannels of polydimethylsi-loxane (PDMS) served as molds that were spontaneously filled owing to capillary forces with suspensions of 0.1 - 40 vol% solid loading. The resulting ceramic lines have a height of about 7 µm and therefore differ from usual ceramic thin film coatings. The capillary filling characteristics were observed under the microscope, the implications of rheology and sus-pension chemistry are discussed and evaluated. Using the same capillaries, even smaller lines (2 - 3 µm width) of powder particles could easily be pre-pared by adjusting only the solid content of the suspensions.

3.1. Introduction

In addition to traditional semiconducting materials, the integration ofceramic and organic materials into microsystems often termed microelectro-mechanical Systems (MEMS) has advanced enormously in recent years. Thefunctionality of such novel microsystems has increased beyond the capabili-ties of electronic circuits and even beyond electromechanical funtionality,including chemical, thermal and optical sensing or actuating. The mostimportant advantages of ceramics in microfabrication are their heat resistance

3.

70 C H A P T E R 3

allowing for high-temperature microdevices and their functional propertieslike piezo- and pyroelectricity and catalytic capabilities. Ceramics like SiCcan play an important role in novel MEMS applications, either as microde-vice substrate for harsh environments[1] or as microrotors.[2] Microdischargereactions were performed inside new monolithic ceramic microdevices.[3]

Functional coatings for instance are applied in micromachined gas sensorswith tin oxide coatings or other catalytically active oxidic layers.[4,5]

Microstructured coatings of functional ceramics may be realised as thinfilms using high-vacuum or sol-gel deposition processes in conjunction withphotolithography. To our knowledge, however, there have been no compara-ble processes with true micrometer resolution in ceramic thick-film process-ing based on colloidal dispersions until recently. The lower resolution limit ofscreen printing typically is about 50 µm. The introduction of ceramic pow-ders in microsystem technology may have significant advantages over thinfilm processes. First, there is much experience in powder synthesis. Exactcontrol of powder grain size, size distribution and morphology is often possi-ble. Second, in cases where dopants are to be controlled, dopant stoichiome-try and its distribution on, respectively inside the grain may be used to fine-tune functional properties. Finally, powder networks with well defined poros-ity can be integrated to exhibit a larger specific surface area.

M I C R O M O L D I N G I N C A P I L L A R I E S 71

Fig. 3–1. Schematic of the micromolding in capillaries process with ceramic suspensions. a) place PDMS mold on substrate. b) apply suspension droplet at end of channels. c) capillary forces fill microchannels within 10-15s. d) wash off excess suspension drop-let, leave to dry. e) remove PDMS mold: microceramic green body structures for sintering.

One method to fabricate powder-based ceramic microstructures is to cast asuspension into appropriate photoresist structures.[6] Alternatively and morecost efficient is the use of soft lithography which was developed as a low costroute to use liquid materials in microfabrication.[7] The use of liquid inor-ganic precursor polymers in soft lithography has already been demonstratedto fabricate microstructured ceramics[8] as well as photonic band gap materi-als and other hierarchically ordered oxides.[9] In soft lithography, micropat-terns are transferred by casting a silicone rubber, poly-dimethylsiloxane(PDMS), against a master structure (see Fig. 3–1). The PDMS is then peeledoff, cut and used as a mold that forms microcapillaries on a substrate whichcan be filled with a liquid. This technique is referred to as micromolding incapillaries (MIMIC). The most striking feature of using PDMS as moldmaterial is, that this elastomeric material readily establishes a reversible con-formal contact on the molecular level on a variety of substrates, thus sealing

e

da

c

b

72 C H A P T E R 3

the capillaries optimally. Additionally, the master structures may be reusedmany times to cast PDMS molds and performing MIMIC itself does notrequire clean room conditions.

We have extended the technique to use colloidal dispersion of ceramicpowders with solid contents of 0.1 to 40 vol%. These methods enabled thefabrication of microstructured ceramic lines with a spatial resolution of10 µm.[10] In another publication we demonstrated how such microceramiclines of tin oxide can be integrated into a functional microdevice and appliedthem as miniaturized gas sensors.[11] A commercially available tin oxide pow-der was used for the present article to demonstrate the capabilities and limita-tions of MIMIC. By varying the solids loading and exploiting the dryingcharacteristics inside the microchannels, continuous lines of only 1-3 µmwidth were obtained. The filling and drying of the dispersion inside the cap-illaries was further characterised by using video microscopy for direct obser-vation of the processes.

3.2. Experimental

Master Structures: The master containing a positive relief of the mold struc-tures were prepared by standard photolithography., AZ 4562 photoresist(Clariant Inc., Wiesbaden, Germany) was spun onto a silicon test wafer at3500 min-1, resulting in a resist thickness of approx. 7 µm. The structureswere transferred by UV illumination through a chromium mask using a SussMA-6 mask aligner. The exposed sample was developed in a Microposit 351(Shipley Inc., Marlborough, MA, USA) - water mixture (ratio 1:3) and couldbe used as master structure several times without further treatment.

PDMS molds: 9 g of Sylgard 184 PDMS prepolymer/catalyst mixture(Dow Corning Inc., Midland, MI, USA) were poured over a master structureand cured for 26 hrs. After peeling off and cutting to pieces containing thepattern accessible from two opposite edges, the PDMS molds were oxygenplasma-treated for 2 min using a 100 W Harrich PDC-32G sterilizer (Har-rich Scientific, Ossinning, NY, USA).

Tin oxide suspension: The tin oxide powder (Cerac Inc., Milwaukee, WI,USA, grain size of d50 = 280 nm as determined by a Microtrac UPA 150 par-ticle sizer) was used as received. In 6.65 g of distilled water (18MOhmcm,

M I C R O M O L D I N G I N C A P I L L A R I E S 73

Millipore), 1.84 g polyacrylic acid sodium salt 2100 (Fluka AG, Buchs, Swit-zerland) was dissolved and 0.06 g ammonia (25% in H2O) added, adjustingthe pH to 9.9. The solution was filtered through a 200 nm pore size teflonfilter and subsequently used as dispersing agent.

33 vol% suspension: to 6.01 g of water and 150 µl of dispersing agent,21.1 g of tin oxide powder were subsequently added, frequently interruptedby ball milling sequences. For other solid loadings, the amounts of powderand dispersing agent were varied accordingly. The suspensions were ball-milled for 20 hrs. The viscosity curves were recorded using a Bohlin CS 50rheometer (Bohlin Inc., Lund, Sweden).

Micromolding in Capillaries: A plasma-treated mold was placed on a sili-con test wafer. 10 µl of tin oxide suspension were dispensed at the entranceof the structures. After the capillary filling, the sample was left to dry at roomtemperature. The dried suspension droplet at the entrance was then removedby pressing the filled PDMS mold tightly to the wafer and rinsing the outsideregion with plenty of water, immediately followed by acetone. After 30 min,the PDMS mold was lifted and the sample was fired at 800°C for 5 hrs.

Video Microscopy: A very thin PDMS stamp was cast by sandwiching themaster structure, PDMS prepolymer mixture and a plastic transparency foilunder a weight of ca. 50 g for curing. Its thickness was defined by usingpieces of silicon wafer (0.38 mm) as spacers. The stamp was demolded care-fully and placed on a silicon wafer substrate with the channel entrances over-lapping the substrate edge by approx. 0.5 mm. The setup was mounted on amicroscope equipped with a video camera using the 50x magnification objec-tive. The tin oxide suspensions were injected through a PE tube from belowand dispensed as close to the capillary entrances as possible.

3.3. Results and Discussion

Microlines of tin oxide ceramics, only 10 µm wide were obtained byMIMIC. A schematic of the process is given in Fig. 3–1. The length of eachline is determined by the characteristics of the capillary filling process andvaried between 0.5 mm (40 vol% suspension) and 5 mm (15 vol%). Sponta-neous filling is thermodynamically driven and occurs when the interfacialfree energies can be minimized by wetting the capillary surface. Kim et al.

74 C H A P T E R 3

have presented a thermodynamic model that describes the filling of a capil-lary with square cross-section.[12, 13] Driving force is the wetting of the chan-nel walls. In order to maximize the driving force of capillary filling, lowcontact angles to PDMS are required. The wetting force is then sufficient toovercome the larger contact angles even of non-wetting substrates. Experi-mentally, we treated the PDMS surface in an RF oxygen plasma to replacesurface methyl groups by hydroxyl groups.[14] The water contact angle of120° is reduced to < 3° after the treatment. The silicon wafer substrate wasnot treated for better wetting with water as its contribution to the free energyis less significant. Although filling of channels is possible in spite the use of ahydrophobic substrate, the possibility of tearing non-adhesive material offthe substrate later in the process upon removing the PDMS mold has to betaken into account.

In the real filling process, the capillary forces have to overcome the viscousdrag of the suspension. For simple cylindrically shaped capillaries, the fillingvelocity as function of filling length is dependent on the volume to surfaceratio of the channel, the contact angles and the viscosity.[15] Solving for fill-ing length z results in a root law

(1)

where r is the hydraulic radius r = V/A (ratio of volume to surface area), γ thesurface tensions for solid-vapour (SV) and solid-liquid (SL), η the viscosity ofthe liquid, and t the time.

However, multiple factors were not considered in this simple model. First,the viscosity of the suspension itself as a non-Newtonian fluid is a function offilling velocity or of the shear rate, respectively. Second, because of continu-ous drying of the suspension during MIMIC, the time before the suspensioncoagulates is limited. This time frame was estimated to be about 30-60 s for a33 vol% suspension. Third, the channel geometry is not cylindrical. In thepresent case, the cross-section of the PDMS capillaries were semi-circular - ashape that resulted from the photoresist profile of the master structures.Fig. 3–2 shows micrographs at different scales of annealed ceramic micro-structures obtained from 33 vol% tin oxide suspensions. Obviously, there is a

zr γSV γSL–( )

2η----------------------------- t⋅=

M I C R O M O L D I N G I N C A P I L L A R I E S 75

statistical distribution of the final microline length due to the various influ-ences just mentioned. Suspensions of higher solids content exhibit muchlarger viscosities than polymer solutions or prepolymer mixtures which werediluted in many cases. Therefore, microchannels of a total length of 5 mmoften may not fill up to their full length. The microchannel geometry with itssharp corners, where PDMS and substrate are joined, also distorts the shapeof the suspension surface during filling. Around those corners, the suspensionseems to travel faster. Therefore, two fork-shaped spikes of 1-2 µm widthemerging from the solid ceramic microstructures are usually observed(Fig. 3–2c). The same behavior also was reported for similar microstructuresfabricated using polymer materials.[12] If the samples are left to dry for a suf-ficiently long time (minimal 30 min), the removal of the PDMS mold bygently tilting it over one edge using tweezers is usually successful, i.e. thestructures are not damaged by the PDMS removal. The structures were thenannealed at 800°C for 5 hrs.

Fig. 3–2. a) top binocular view on annealed microstructured lines of tin oxide on a silicon wafer substrate as obtained by spontaneous capillary filling from a 33% suspension. b) SEM close-up of a line showing its porous microstruc-ture. c) structures at the end of the microlines. The fork-like spikes origi-nate from the wetting characteristics of suspensions inside non-circular microchannels.

76 C H A P T E R 3

The lines consist of an annealed powder network exhibiting an estimatedporosity of 40-50%[10] with high specific surface which was intended for theapplication of such microstructures as miniaturized semiconducting gas sen-sors. The effect of solid loading was evaluated by preparing suspensions ofsolid contents ranging from 40% to 0.1%. We distinguished between highsolid loadings > 15 vol% and low solid loadings < 1 vol%. For high solidloadings, viscosity curves, the average filling length and an evaluation of fill-ing length versus viscosity are shown in Fig. 3–3. When preparing suspen-sions for MIMIC, the viscosity needs to be minimized. Tin oxide powder wasin the present case sterically stabilized using a ammonium polyacrylate liq-uidifying agent, while the pH of the suspension was set to 8. Using this com-positions, it was possible to obtain viscosities lower than 100 mPas (100s-1)up to 33 vol%, which is reasonably close to water (1 mPas, see Fig. 3–3a).Even these comparably low values for the viscosity lead to a viscous drag inthe range of the capillary forces as stated in equation (1). Fig. 3–3b shows themean final lengths of 10 µm wide lines versus the solid loading of the suspen-sion. The error bars mark the standard deviation of 20 lines evaluated foreach suspension. As expected, the filling length decreases with higher solidloadings. Only the 15 vol% suspension reached the other end after approx.5 mm and filled the available channel length of the mold fully. On the otherhand, the scattering of the lengths at high solid loadings becomes smaller, theMIMIC process more reproducible. It is important to note that these resultsapply only to the spontaneous filling process and to this powder. Other pow-ders will result in different rheology behavior and hence, in different fillingcharacteristics. As other experiments showed, MIMIC can be performedusing a variety of different powders (Al2O3, ZrO2) provided that the grainsize is sufficiently small (roughly 1/10 of the capillary diameter) and that low-viscosity suspensions can be prepared. No external aids to further extend thefilling length were applied. Possible methods could consist of slowing downthe drying of the suspension by cooling the suspension to lower temperaturesor of performing MIMIC under ultrasonic agitation. Alternatively, the elasto-meric properties of the PDMS polymer can be of use to keep the filling pro-cess running by gently compressing and releasing the PDMS mold.Occasionally, it was possible to fill a capillary of 5 mm length with 33% sus-

M I C R O M O L D I N G I N C A P I L L A R I E S 77

pension using this PDMS compressing method. However, it proved difficultto quantify the impact of such length-enhancing methods. A maximum vis-cosity of 0.1 - 0.2 Pa s was estimated from the results shown in Fig. 3–3c toobtain filling lenghts of at least 1 mm. We consider a filling length in therange of 1 mm well suited for most prototype MEMS applications. Qualita-tive agreement with the model from Eq. (1) was observed. The real processseems to be even more sensitive to viscosity as the filling length decreasesmore rapidly as the fitted η-1/2 curve. Based on observed initial filing veloci-ties of 60-75 µm s-1 and the dimensions of the capillary, a shear rate of about15 s-1 was calculated for this MIMIC process. Therefore, the viscosity valuesevaluated in Fig. 3–3c were taken at 15 s-1. Video microscopy on capillaryfilling of 33% suspensions also revealed that the filling is a discontinuousprocess. The running front of the suspension stops for a few ms after everyentering 40 - 100 µm before continuing to advance inside the capillary. Withtime, the number of phases of movement becomes more and more scarceuntil the suspension begins to coagulate due to drying.

78 C H A P T E R 3

Fig. 3–3. a) viscosity curves measured on the tin oxide suspensions steri-cally stabilized with a poly-acrylate additive. b) final filling length of 10 mm wide PDMS capillaries as function of the solids loading of the tin oxide suspensions used. The value of 15 vol% suspen-sions is bracketed because the approx. 5 mm capillary length was filled up to its end, whereas all other suspensions did not fill the capillaries completely. c) plot of filling length versus viscosities taken at shear rates of 15 and 100 s-1.

a

b

c

M I C R O M O L D I N G I N C A P I L L A R I E S 79

Fig. 3–4. Scanning electron microscope cross-sections of microlines prepared by snapping the silicon wafer substrates. The samples were imaged without conductive caotings.

A second, equally important issue is the reproduction quality of the ceramicmicrolines. Although the 15 vol% suspension filled the channels completely,the distribution of powder after drying was very uneven. Many lines evenexhibited gaps, where not enough powder was available to fully coat the sub-strate surface. Again, video microscopy served as an important tool to gain abetter understanding of the process. As the solid loading is reduced to15 vol% and lower, the suspension inside the capillary takes much longer todry. Eventually, the suspension remainder droplet outside the PDMS molddries before the suspension inside the capillaries. In those cases, the liquidphase inside the capillaries was observed to be drawn out of the capillary atvery high speed due to capillary forces in the porous network of the drieddroplet outside the PDMS. The stabilized fraction of particles is drawn alonguntil it hits areas of precoagulated particles where they block the channelquickly, leading to an uneven distribution of powder inside the capillary. The

80 C H A P T E R 3

receding suspension front again becomes fork-like shaped. Therefore, thecorners of the capillaries dry last and thereby the remaining particles arealigned along the edges of the capillary. Such powder movements during dry-ing were observed up to 25 vol% suspensions, although the impact on thefinal microstructure with 25% suspensions was not as pronounced. In thatcase, it only textured the surface of the line. Having discussed width and fill-ing length in suspension-based MIMIC, the resulting structure height andcross-sections of the microlines are assessed in Fig. 3–4. The cross-sectionswere prepared by scoring and snapping the silicon wafer substrates perpen-dicular to the microlines. Only 40% and 33% suspensions were able to fillthe full height of the capillaries of 6.5 -7 µm, while the 15% microline has atypical thickness of only 2 µm. The bell-shaped outline is attributed to dry-ing shrinkage. This may also be of advantage for the removal of the PDMSmold, since the green body is assumed to break contact with the PDMS sur-face. For the application as miniaturized gas sensors, the inclined side wallsalso facilitated the electrical contacting by sputtering Pt metal contactsdirectly onto the microlines. Should a steeper side wall profile be required, analternative microfabrication approach of casting ceramic suspensions directlyinto photo resist structures can be taken.[6,16]

MIMIC with suspensions of very low solid loading can produce evensmaller structures than the capillary. Lines of only 2 - 3 µm width were pro-duced by exploiting the spikes of the drying suspension front along thePDMS-substrate corners (see Fig. 3–5). An appropriate solid content thatleads to continuous lines was found at 1%. Lower solid loadings (e.g., 0.1%)resulted in unconnected dots of powder, higher solid loadings to unspecifi-cally coated areas within the capillaries. Without changing the capillary size,line structures a factor of 5 smaller were obtained. A 1% suspension wasobserved to remain liquid inside the channels up to 30 min before drying offfast. In previous publications we have exploited a similar process for thealignment of V2O5 nanotubes by MIMIC.[17]

M I C R O M O L D I N G I N C A P I L L A R I E S 81

Fig. 3–5. Very small double microlines of 1-2 µm width produced using a 1% tin oxide suspension. Higher solid contents lead to powder depositions on the whole substrate surface inside the microchan-nels, lower solid loadings to disconnected lines. The inside edge of a double line is clearly resolved, whereas the suspension pene-trated the interface between PDMS mold and substrate some-what, leading to rough outside edges.

3.4. Summary

A low cost method to introduce new line-shaped ceramic coatings made fromtin oxide powders was presented. Unlike traditional thin films, these struc-tures feature an aspect ratio about 1 and, depending on the powder andannealing conditions used, may exhibit a controlled porosity. The full rangeof available ceramic powders is applicable, provided that the powder can bestabilized in a colloidal suspension to yield sufficiently low viscosities. Usingmicromolding in capillaries, ceramic components with typical cross-sectionsof 3 - 10 x 10 µm2 and lengths up to 5 mm are easily reproduced. However,the method lacks exact control on the filling length. MIMIC with ceramicsuspensions is a rapid prototyping method before resorting to more sophisti-

82 C H A P T E R 3

cated microfabrication technology. Possible fields of application include func-tional ceramic coatings in sensors, or as structural elements, possibly as heatresistant spacers. Such microstructures may also be of service in ceramicmaterials science, creating addressable microstructures, e.g. where singlegrains transforming during a sintering process may be found and reanalyzedafterward. Different ceramic powders deposited close to each other may besubjected to identical treatment conditions (thermal history, etc.) due to theirproximity.

M I C R O M O L D I N G I N C A P I L L A R I E S 83

3.5. References

[1] M. Mehregany, C. A. Zorman, N. Rajan, C. Hung Wu, Silicon Carbide MEMS for Harsh Environments, Proc. IEEE, 1998, 86, 8, 1594-1610.

[2] J. F. Li, S. Sugimoto, S. Tanaka, M. Esashi, R. Watanabe, Manufacturing silicon carbide microrotors by reactive hot isostatic pressing within micromachined silicon molds, J. Am. Ceram. Soc., 2002, 85, 1, 261-263.

[3] B. A. Vojak, S. J. Park, C. J. Wagner, J. G. Eden, R. Koripella, J. Burdon, F. Zen-hausern, D. L. Wilcox, Multistage, monolithic ceramic microdischarge device hav-ing an active length of similar to 0.27 mm, Appl. Phys. Lett., 2001, 78, 10, 1340-1342.

[4] H. L. Tuller, Advanced Sensor Technology Based on Tin Oxide Thin Film - MEMS Integration, J. Electroceram., 2000, 4, 2, 415-425.

[5] N. Barsan, M. Schweizer-Berberich, W. Göpel, Fundamental and practical aspects in the design of nanoscaled SnO2 gas sensors: a status report, Fresenius J. Anal. Chem., 1999, 365, 287-304.

[6] U. Schönholzer, R. Hummel, L. J. Gauckler, Microfabrication of Ceramics by Fill-ing of Photoresist Molds, Adv. Mater., 2000, 12, 17, 1261-1263.

[7] Y. Xia, G. M. Whitesides, Soft Lithography, Angew. Chem. Int. Ed., 1998, 37, 550-575.

[8] H. Yang, P. Deschatelets, S. T. Brittain, G. M. Whitesides, Fabrication of high per-formance ceramic microstructures from a polymeric precursor using soft lithogra-phy, Adv. Mater., 2001, 13, 1, 54.

[9] P. Yang, T. Deng, D. Zhao, P. Feng, D. Pine, B. F. Chmelka, G. M. Whitesides, G. D. Stucky, Hierarchically Ordered Oxides, Science, 1998, 282, 2244.

[10] M. Heule, L. Meier, L. J. Gauckler, Micropatterning of Ceramics on Substrates towards Gas Sensing Applications, Mat. Res. Soc. Symp. Proc., 2001, Vol. 657, EE9.4.

[11] M. Heule, L. J. Gauckler, Gas Sensors Fabricated from Ceramic Suspensions by Micromolding in Capillaries, Adv. Mater., 2001, 13, 23, 1790.

[12] E. Kim, Imbibition and Flow of Wetting Liquids in Noncircular Capillaries, J. Phys. Chem. B, 1997, 101, 855-863.

[13] E. Kim., Y. Xia, G. M. Whitesides, Polymer microstructures formed by moulding in capillaries, Nature, 1995, 376, 581-584.

[14] M. J. Owen, "Surface chemistry and applications" in “Siloxane Polymers”; edited by S. J. Clarson, J. A. Semlyen, PTR Prentice Hall, Englewood Cliffs, NJ, 1993, 309-372.

[15] E. Kim, Y. Xia, G. M. Whitesides, Micromolding in Capillaries: Applications in Materials Science, J. Am. Chem. Soc., 1996, 118, 5722-5731.

[16] M. Heule, L. J. Gauckler, Microfabrication of Ceramics based on Colloidal Sus-pensions and Photoresist Masks, J. Photopolym. Sci. Technol., 2001, 14, 3, 449.

[17] H. J. Muhr, F. Krumeich, U. P. Schonholzer, F. Bieri, M. Niederberger, L. J. Gauckler, R. Nesper, Vanadium oxide nanotubes - A new flexible vanadate nanophase, Adv. Mater., 2000, 12, 3, 231.

84 C H A P T E R 3

85

M. Heule, U. P. Schönholzer, L. J. Gauckler,submitted to Adv. Mater, 2003.

Patterning Colloidal Suspensions by Selective Wet-ting of Microcontact-Printed Surfaces

Micropatterns of ceramic powders can be obtained by selective wetting of microcontact-printed surfaces. A large wetting contrast between hydro-philic micropatterns and hydrophobic areas was created. Aqeous colloidal dispersions of aluminum oxide and tin oxide adhered only to the hydro-philic micropatterns whereas they were repelled from the hydrophobic areas in a simple dip coating process. We examined two molecular ink/sub-strate systems: thiol self-assembled monolayers (SAM) on gold and octade-cyltrichlorosilane (OTS) SAM on silicon wafer substrates. Corresponding contact angles obtained under varying printing conditions are presented. The chemical compositions of the printed layers were characterized by ToF-SIMS mass spectrometry. The thiol-gold SAM readily forms in micro-contact printing whereas the OTS layer also contains a significant amount of PDMS residues. However, printing and selective wetting could be car-ried out successfully on both ink/substrate systems. Ceramic micropatterns with a resolution of 5 µm are shown.

4.1. Introduction

Microcontact Printing (µCP) is an extremely versatile alternative to standardphotolithography for direct micropatterning of a variety of substrates. It wasintroduced as one of a whole set of techniques for alternate microfabrication(Soft Lithography) using elastomeric molds for pattern transfer.[1] Elasto-meric stamps are easily prepared by casting a prepolymer of polydimethylsi-loxane (PDMS) against the topography of a master structure which

4.

86 C H A P T E R 4

themselves may be prepared by photolithography or other microfabricationmethods. The stamps containing recessed microstructures are inked and sub-sequently brought into contact with substrates to transfer molecules inmicropatterns by a simple printing process (Fig. 4–1a). The best character-ized ink system is the printing of alkylthioles that form self-assembled mono-layers (SAM) on gold surfaces.[2,3] The SAM may be used as an ultimativelythin mask for etching the gold layer selectively.[4,5] This has been demon-strated by using µCP for creating microelectrodes for cyclic voltammetry.[6]

The achievable resolution may be in the submicron range, even on largeareas.[7] Substrates do not necessarily need to be flat, as the work of Jackmanet al. demonstrates.[8] Other printable substances include an enormous rangeof both ink and different substrates, for instance the printing of Pd catalystsfor electroless deposition of patterned metal films,[7] zeolite layers[9] or bio-functional layers like patterned protein repellent polylysine-polyethylenegly-col co-polymers for cell adhesion studies.[10,11] Protein microarrays have alsobeen presented.[12] Since µCP can be carried out without clean room equip-ment, the method is spreading quickly among various disciplines. One of themost striking features of µCP is the possibility to scale up the process for con-tinuous mass fabrication since the process is the microfabrication equivalentof printing on paper. The group of Michel at IBM Corporate Research hasundertaken first steps in that direction,[13,14] thereby synthesizing more sta-ble PDMS-analog elastomers than the commercially available standardblend.[15]

M I C R O C O N T A C T P R I N T I N G 87

Fig. 4–1. a) principle of microcontact printing. (1) PDMS stamp is inked with a SAM-forming molecule in solution and dried (2). The stamp is brought into contact with a substrate (3), transferring its pattern by forming the SAM (4). Substrates are then coated with colloidal suspensions by dip coating, yielding a surface with micropat-terned thick films (5). b) reactions of SAM formation with hexade-canethiol (HDT) on gold and octadecyltrichlorosilane (OTS) on SiO2 surfaces.

In this article, we propose to supplement µCP with a scheme for the micro-fabrication of ceramic thick films. We use µCP for generating contrastinghydrophilic and hydrophobic areas which are then selectively wetted by a col-loidal suspension to form micropatterns of ceramics. The article is part of aseries in which other soft lithographic techniques have been used to fabricatesmall elements of ceramic powders.[16-20] Powders may offer several advan-

PDMS stamp

1.

2.

Substrate

3.

4.

5.

a

b

Au+

S

H

C16

H33

H2O+

Si

OH

Si-Wafer

O Si

O

Si

C18H37

+ 3 HCl( )

n

Alkylthiol-Au SAM formation

Alkylsiloxane-SiO2 SAM formation

S

Au

C16

H33

+ 1/2 H2

HDT

Si Cl

ClCl

C18H37

OTS

88 C H A P T E R 4

tages over standard thin film ceramics, e.g. higher sensitivity and controlledporosity for gas sensing applications.[21,22] We examine two ink/substratesystems, the thiol-gold and the metal-less alkanetrichlorosilane chemistrywhich allows to print directly onto bare silicon wafer substrates, see Fig. 4–1b. The layers obtained by µCP have been characterized by water contactangle measurements and more sensitively by ToF-SIMS (time-of-flight sec-ondary ion mass spectrometry).

4.2. Experimental

The masters containing a positive relief of the mold structures were preparedby standard photolithography using AZ4562 positive resist (Clariant GmbH,Wiesbaden, Germany). 9 g of Sylgard 184 PDMS prepolymer/catalyst mix-ture (Dow Corning Inc., Midland, MI, USA) were poured over a masterstructure and cured for 26 hrs.

The substrates were silicon wafer pieces with a freshly evaporated layer ofgold prepared by thermal evaporation in a Balzers MED 020 system at apressure of 1.7 ·10-5 mbar, thickness ranging from 10.3 to 33.7 nm, includ-ing a layer of Cr (thickness 6.7 nm) in between to promote adhesion. Fordirect printing of octadecyltrichlorosilanes, bare Si wafer pieces were cut,rinsed with water (18 MΩ cm, Millipore), cleaned in an ultrasonic bath andsubsequently treated in an oxygen plasma sterilizer for 2 min (Harrick PDC-32G).

µCP on Au layers. The PDMS stamp was inked for 30 min in 25 mL of1-hexadecanethiol solution (Fluka AG, Buchs, Switzerland, used as received)in ethanol (38.2 µL, 1.25 x 10-4 mol, 5 mM). After this time the stamp wastaken out of the solution and dried with a stream of nitrogen. Then it wasplaced on the substrate for 20 s before being removed. The printed substratewas immersed in a 25 mL solution of cysteamine hydrochlorideHS(CH2)2NH2HCl (Fluka) in ethanol (0.0142 g, 1.25 x 10-4 mol, 5 mM)for 30 min, after which it was removed, washed in ethanol and dried with astream of nitrogen.

M I C R O C O N T A C T P R I N T I N G 89

µCP on Si. A 5 mM solution of Octadecyltrichlorosilane OTS (Fluka) ineither hexane or ethanol was prepared just minutes before printing. Micro-contact printing was carried out the same way as detailed above, with a con-tact time of 60 s. The substrates were heated to 60°C for 15 min and thenrinsed with the solvent in use.

Al2O3 suspensions (190 nm median particle size, 45 vol. %; TaimeiIndustries, Tokyo, Japan, TM-DAR 2831). The Al2O3 suspension (27.87mL) was prepared in pure water, NH4Cl (0.0745g, 1.39 mmol, 0.05 M) andHCl (2 M, Titrisol, Fluka). An arbitrary amount of HCl was used here toreduce the pH of the suspension into the region 4-5. This suspension wasball milled overnight before use.

SnO2 suspensions (220 nm median particle size, 33 vol%; Cerac Inc.,Milwaukee, WI,USA). 21.1 g of SnO2 were added to 6.01 g of water con-taining 150 µL of dispersing agent which was prepared by mixing 6.65gwater, 1.84 g polyacrylic acid sodium salt 2100 (Fluka) and 0.06g NH3, 25%in H2O). Again, the suspensions were homogenized by ball milling sequencesthroughout the powder adding procedure.

Coating the printed substrates. The substrates were coated with suspen-sion either by dip-coating or by dropping the suspension onto a tilted sub-strate (tilt angle approx. 70°). Sintering was done by heating the samples to800°C for 5 hrs for SnO2, resp. 1100°C for 2 hrs for Al2O3 coatings.

Surface characterisation. The static contact angle of water was deter-mined on all samples. ToF-SIMS mass spectrometry was carried out in aPHI 7200 analyzer using a Cs+ primary ion beam with an incident energy of8 keV. Samples were microcontact-printed not more than 3 hrs before mea-surement using blank PDMS stamps without microstructures. Referencesamples were prepared by immersion in the ink solutions over night for thefull SAM coverage references, and treating bare Si, resp. Au-coated piecesonly with oxygen plasma for the blank reference samples. The ion dose wasbelow the static limit (<1.0 · 1012 ions/cm2). The primary ion beam wasscanned over an area of 200 by 200 µm2. Aqcuisition pulses were of 1.25 nsduration with an acquisition time of 1.7 min/spectrum. Typical m/∆m reso-

90 C H A P T E R 4

lutions obtained ranged from 3000 to 5000. The spectra were fine-calibratedwith the PHI Tofpak software using the exact masses of a standard set ofsmall ions.

4.3. Results and Discussion

The alkanethiol and alkanetrichlorosilane molecules shown in Fig. 4–1b areused to create the hydrophobic surfaces. In the case of gold-thiols, theunprinted regions had to be backfilled with another thiol, cysteamine hydro-chloride HS(CH2)2NH2 HCl, forming the hydrophilic layer with its ammo-nia head groups. The substrates were then dip-coated in 45 vol% aluminasuspensions, resulting in the coating of the line pattern after drying as shownin Fig. 4–2a (dark lines are ceramic, gold is bright). AFM cross-sections ofthe resulting ceramic coatings are displayed in the insets. The rounded shapesare a combined result of the contact angle of wetting, the surface tension ofthe suspension and the drying shrinkage. For a given micropattern geometry,the height of the coating can be influenced by the suspension properties.

The addition of NH4Cl to the suspension was crucial. Without salt, thesuspension did not selectively wet the patterned substrate, only a dense, con-tinuous layer was formed. When added to pure water, salt increases the sur-face tension of the solution slighty. In order to get a rough estimate of thechange of the surface tension of alumina suspensions by the NH4Cl addition,the weight of suspension droplets generated with a pipette was measured.The surface tension σ of a liquid is related to the weight of a droplet Vρg andthe radius of the pipette tip rc.[23]

(1)Vρg 2πrcσ=

M I C R O C O N T A C T P R I N T I N G 91

Fig. 4–2. a) micropatterned ceramic lines of 20 µm width on gold sub-strates. b) resolution test structure of triangles that narrow down to form 5 µm wide lines.

The surface tension of an alumina suspension is significantly lowered by thesalt addition from σ = 8.8 · 10-2 N m-1 to 3.9 · 10-2 N m-1. It seems that alower surface tension is favorable for the wetting process. The contact anglesof water on the surfaces prepared in this work are summarized in Table 4–1.In the thiol-gold system, a contact angle difference of alkanethiol/cysteamineSAM layers of approx. 65° is reached. This contrast is sufficient for selectivewetting of colloidal suspensions.

92 C H A P T E R 4

In Fig. 4–2b, another ceramic coating of a triangular test pattern demonstrat-ing the resolution capabilities of the process is shown. This sample was pre-pared by droplet deposition. The triangles narrow down to lines of 5 µmwidth which are clearly resolved with a contiguous layer of a few particles.However, the edges of the triangles are not always resolved sharply. Whenlooking close at the hydrophobic/hydrophylic boundary, fractal-like or den-dritic structures are observed. This is probably due to the backfilling processduring which cysteamines may have partially substituted alkanethiols.

In the case of the alkanetrichlorosilane approach, a backfilling is notneeded since the silicon surface is very hydrophilic after the oxygen plasmatreatment. However, the surface chemistry is more complicated. The SiCl3functional group is much more reactive than SH and can readily polymerizein presence of small amounts of water like air humidity. And as indicated inFig. 4–1b, water is necessary for covalently linking the trichlorosilane headgroup to silicon oxide. Usually, such OTS layers are used for friction reduc-tion in silicon-based MEMS devices. They are coated with OTS from solu-tion in a dry solvent such as hexane.[24,25] The heating step to 60°C wassuggested for completing the cross-linking reaction. It is suggested that thesurfaces under ambient conditions contain enough adsorbed water for cross-linking OTS to a SAM. Jun et al. circumvent this problem by treating the sil-icon surface with chlorine, then printing octadecanol.[26] The main draw-back of this approach is the need for an inert atmosphere due to the reactivityof Si-Cl species. The PDMS stamp is temporarily dried in the µCP process.Interfering reactions of water from PDMS surfaces or air with OTS mayoccur before SAM formation on the surface. Therefore, it is not clearwhether a well ordered SAM is formed under typical microcontact printingconditions. Experiments in which PDMS was immersed for a few minutes inhexane showed that PDMS is damaged by extraction of components by thesolvent. PDMS looses its ability to spontaneously seal smooth surfaces andbecomes white-opaque and brittle. Therefore, it was decided to use ethanolas solvent for OTS. A contrast in water contact angle of up to 97° could becreated in spite of these suspected side reactions (Table 4–1). Ceramic micro-coatings of 33% tin oxide suspensions could be formed successfully by dip-

M I C R O C O N T A C T P R I N T I N G 93

coating (Fig. 4–3a), respectively droplet coating (Fig. 4–3b). When printingstructures in the 100 µm range, the suspensions dry unevenly and form shal-low holes in the center of the coated surfaces.

Table 4–1. Summary of water contact angles θ. which were measured in static mode. Where available, values are presented as range over several samples.

θwater PDMSAu Substrate

Si Substrate

PDMS 110°

PDMS O2 plasma-treated

< 2°

Au layer 76°

µCP Au-HDT SAM 108°

Au-HDT from EtOH solution

105°

µCP Au-Cysteamine HCl 43-46°

Si as received 33-36°

Si O2 plasma-treated < 2°

µCP OTS/hexane ink 95-110°

µCP OTS/ethanol ink 99-104°

OTS from hexane solu-tion

99°

µCP PDMS dry 60-67°

94 C H A P T E R 4

Fig. 4–3. a) cross structure of tin oxide formed by the OTS/Si µCP approach. b) 20 µm tin oxide line pattern coated on Si wafers.

Sintering of the microcoatings was performed up to 1100°C without crack-ing the layers. However, the oxidation of silicon has to be taken into account.Under wet oxidation conditions, the ceramic layer can partially be embeddedin a SiO2 layer. If the Au layer is 100 nm, delamination of the ceramic micro-

M I C R O C O N T A C T P R I N T I N G 95

lines may occur due to the gold coalescing into droplets. This delaminationcan be avoided using thinner gold films around 30 nm. This effect could alsobe used deliberately for separating ceramic parts from their substrate support.In order to learn more about the composition of these hydrophobic layers,ToF-SIMS measurements were performed on micro-contact-printed siliconand gold surfaces. Since alkanes do not form many stable anions whereas Auand Sulfur-containing species do, the negative mode mass spectra yieldedmore information. ToF-SIMS measurement data of alkanethiol SAMs on Auhave been published before.[27] Mass spectra of µCP HDT-Au samples arecompared to a reference SAM in Fig. 4–4a. A good match was obtained,there are no obvious differences. The formation of various AuS-fragmentsproved very useful for direct confirmation of a sulfur-gold chemical bond.Minor amounts of PDMS were also detected.

Microcontact-printed OTS layers on Si are not as easily identified as inthe case of Au-S SAMs. The layer is of similar composition as the substrate.Si possibly originates from the wafer, its native SiO2 layer, from PDMS andfrom OTS. Cl is expected to evaporate as HCl during the coating process andgeneral alkane signals are too unspecific for identification of OTS. Therefore,the only mean of unambiguously establishing the presence of an OTS-layer isthe detection of intact octadecyltrioxysilane species, e.g. m/z = 329(SiO3(CH2)17CH3)- or m/z = 331 (H2SiO3(CH2)17CH3)-, which could bedetected (see Fig. 4–4b). Such fragments have also been identified by Hous-siau et al. who examined OTS SAM on aluminum metal surfaces.[28] How-ever, the mass spectrum of the microcontact-printed sample strongly matchesthe PDMS reference spectrum. The di- and trimonomeric PDMS fragmentseries m/z = 119 (CH3Si2O3)-, 133 (C2H5Si2O3)-, 149 (C3H9Si2O3)-, 223(C5H15Si3O4)- are prominently present including Si isotope signals and frag-ments lacking H-atoms.

ToF-SIMS analysis confirmed the presence of a thiol-gold SAM with onlyminor amounts of PDMS contamination. In the case of transferring octade-cyltrichlorosilanes by micro-contact printing, there seems to be a consider-able amount of PDMS residues present. The relative amount of detectedOTS fragments decreased by approximately 10 times, compared to a OTSreference surface. However, this differences in chemical composition of the

96 C H A P T E R 4

obtained layers do not seem to affect the macroscopic effect of selective wet-ting adversely. For further improvements of the method, however, these find-ings will have to be taken into account carefully. A possible explanation forthese differences to the Au-HDT system could be that OTS is able to dissolvesome PDMS species which are deposited upon printing.

Fig. 4–4. a) negative mode ToF-SIMS mass spectra of microcontact-printed HDT on Au at various scales compared to their respective reference samples (below). b) ToF-SIMS results from the OTS-SiO2 system.

190 200 210 220 230 240 250 260 270 280 2900

2000

4000

6000

8000

m/z

Co

un

ts

MH15509.SUR

390 395 400 405 410 415 420 425 4300

100

200

300

m/z

MH15509.SUR

190 200 210 220 230 240 250 260 270 280 2900

2000

4000

6000

8000

m/z

Co

un

ts

MH15507.SUR

Au-

196.97AuS

-

228.94AuSCH2CH2

-

256.97

AuS2-

260.91

390 395 400 405 410 415 420 425 4300

100

200

300

m/z

MH15507.SUR

Au2-

393.93Au2S

-

425.91

Au-SAM reference Au-SAM reference

a

PDMS reference

100 120 140 160 180 200 220 2400

1000

2000

3000

m/z

Co

un

ts

MH19006C.SUR

100 120 140 160 180 200 220 2400

200

400

600

800

1000

m/z

Co

un

ts

MH19010C.SUR

327 328 329 330 331 332 333 334 335 336 3370

5

10

15

20

25

m/z

Co

un

ts

MH19010C.SUR

OTS µ-CP samples

327 328 329 330 331 332 333 334 335 336 3370

50

100

150

200

250

m/z

Co

un

ts

MH15533C.SUR

OTS reference

b

Au-SAM µ-CP samples

M I C R O C O N T A C T P R I N T I N G 97

4.4. Summary

Microcontact printing with colloidal suspensions of ceramic powders can besuccessfully performed using thiol-gold SAM and OTS-PDMS-silicon oxidechemistries. Hydrophobic/hydrophilic contrasts of 62° to 97° direct the wet-ting of the suspension, resulting in ceramic micropatterns with a resolutionof 5 µm. The process is versatile for different powder dispersions, as has beenshown using alumina and tin oxide colloidal suspensions. This elegant andcost-effective procedure also has potential for wafer-level fabrication.

98 C H A P T E R 4

4.5. References

[1] Y. Xia, G. M. Whitesides, Soft Lithography, Angew. Chem. Int. Ed., 1998, 37, 550-575.

[2] E. Delamarche, B. Michel, H. A. Biebuyck, C. Gerber, Golden Interfaces: The Sur-face of Self-Assembled Monolayers, Adv. Mater., 1996, 8, 9, 719-729.

[3] E. Delamarche, H. Schmid, A. Bietsch, N. B. Larsen, H. Rothuizen, B. Michel, H. Biebuyck, Transport Mechanisms of Alkanethiols during Microcontact Printing on Gold, J. Phys. Chem. B, 1998, 102, 3324-3334.

[4] E. Delamarche, M. Geissler, H. Wolf, B. Michel, Positive microcontact printing, J. Am. Chem. Soc., 2002, 124, 15, 3834-3835.

[5] M. Geissler, H. Schmid, A. Bietsch, B. Michel, E. Delamarche, Defect-tolerant and directional wet-etch systems for using monolayers as resists, Langmuir, 2002, 18, 6, 2374-2377.

[6] B. A. Grzybowski, R. Haag, N. Bowden, G. M. Whitesides, Generation of Micrometer-Sized Patterns for Microanalytical Applications Using a Laser Direct-Write Method and Microcontact Printing, Anal. Chem, 1998, 70, 4645-4652.

[7] T. Burgin, V. E. Choong, G. Maracas, Large Area Submicrometer Contact Printing Using a Contact Aligner, Langmuir, 2000, 16, 5371-5375.

[8] R. J. Jackman, S. T. Brittain, G. M. Whitesides, Fabrication of Three-Dimensional Microstructures by Electrochemically Welding Structures formed by Microcontact Printing on Planar and curved Substrates, IEEE J. MEMS, 1998, 7, 2, 261-265.

[9] K. Ha, Y. J. Lee, D. Y. Jung, J. H. Lee, K. B. Yoon, Micropatterning of oriented zeolite monolayers on glass by covalent linkage, Adv. Mater., 2000, 12, 21, 1614.

[10] G. Csucs, R. Michel, J. W. Lussi, M. Textor, G. Danuser, Microcontact Printing of novel co-polymers in combination with proteins for cell-biological applications, Biomaterials, 2002, (in press).

[11] N. P. Huang, R. Michel, J. Voros, M. Textor, R. Hofer, A. Rossi, D. L. Elbert, J. A. Hubbell, N. D. Spencer, Poly(L-lysine)-g-poly(ethylene glycol) Layers on Metal Oxide Surfaces: Surface-Analytical Characterization and Resistance to Serum and Fibrinogen Adsorption, Langmuir, 2001, 17, 489-498.

[12] J. P. Renault, A. Bernard, D. Juncker, B. Michel, H. R. Bosshard, E. Delamarche, Fabricating microarrays of functional proteins using affinity contact printing, Angew. Chem.-Int. Edit., 2002, 41, 13, 2320-2323.

[13] B. Michel, A. Bernard, A. Bietsch, E. Delamarche, M. Geissler, D. Juncker, H. Kind, J. P. Renault, H. Rothuizen, H. Schmid, P. Schmidt-Winkel, R. Stutz, H. Wolf, Printing meets lithography: Soft approaches to high-resolution printing, IBM J. Res. Dev., 2001, 45, 5, 697-719.

[14] A. Bietsch, B. Michel, Conformal contact and pattern stability of stamps used for soft lithography, J. Appl. Phys., 2000, 88, 4310-4318.

[15] H. Schmid, B. Michel, Siloxane Polymers for High-Resolution, High-Accuracy Soft Lithography, Macromolecules, 2000, 33, 8, 3042-3049.

[16] M. Heule, L. Meier, L. J. Gauckler, Micropatterning of Ceramics on Substrates towards Gas Sensing Applications, Mat. Res. Soc. Symp. Proc., 2001, Vol. 657, EE9.4.

[17] M. Heule, L. J. Gauckler, Microfabrication of Ceramics based on Colloidal Sus-pensions and Photoresist Masks, J. Photopolym. Sci. Technol., 2001, 14, 3, 449.

M I C R O C O N T A C T P R I N T I N G 99

[18] M. Heule, J. Schell, L. J. Gauckler, Powder-based Tin Oxide Micro-Components on Silicon Substrates fabricated by Micromolding in Capillaries, J. Am. Ceram. Soc, 2002, (in press).

[19] U. Schönholzer, R. Hummel, L. J. Gauckler, Microfabrication of Ceramics by Fill-ing of Photoresist Molds, Adv. Mater., 2000, 12, 17, 1261-1263.

[20] U. P. Schonholzer, N. Stutzmann, T. A. Tervoort, P. Smith, L. J. Gauckler, Micro-patterned ceramics by casting into polymer molds, J. Am. Ceram. Soc., 2002, 85, 7, 1885-1887.

[21] M. Heule, L. J. Gauckler, Miniaturised Arrays of Tin Oxide Gas Sensors on Single Microhotplate Substrates Fabricated by Micromolding in Capillaries, Sens. Actuator B, 2002, (in press).

[22] M. Heule, L. J. Gauckler, Gas Sensors Fabricated from Ceramic Suspensions by Micromolding in Capillaries, Adv. Mater., 2001, 13, 23, 1790.

[23] M. Vogel, "Physik", 20 ed., edited by C. Gerthsen, Springer, Berlin, 1995, 101.[24] T. Ito, M. Namba, P. Bühlmann, Y. Umezawa, Modification of Silicon Nitride Tips

with Trichlorosilane Self-Assembled Monolayers (SAMs) for Chemical Force Microscopy, Langmuir, 1997, 13, 4323-4332.

[25] U. Srinivasan, M. R. Houston, R. T. Howe, R. Maboudian, Alkyltrichlorosilane-Based Self-Assembled Monolayer Films for Stiction Reduction in Silicon Microma-chines, IEEE J. MEMS, 1998, 7, 2, 252-260.

[26] Y. Jun, D. Le, X. Y. Zhu, Microcontact printing directly on the silicon surface, Langmuir, 2002, 18, 9, 3415-3417.

[27] D. A. Hutt, G. J. Leggett, Static Secondary Ion Mass Spectrometry studies of self-assembled monolayers: electron beam degradation of alkanethiols on gold, J. Mater. Chem., 1999, 9, 923-928.

[28] L. Houssiau, P. Bertrand, Tof-SIMS study of organosilane self-assembly on alumin-ium surfaces, Appl. Surf. Sc., 2001, 175-176, 351-356.

100 C H A P T E R 4

101

M. Heule, L. J. Gauckler,J. Photopolym. Sci. Technol., 2001, 14 (3), 449.

Casting of Suspensions into Photoresist Structures

Thick film ceramic microstructures with cross sectional areas of 5 by 10 µm2 were fabricated by casting aqueous colloidal suspensions of fine ceramic powders into photoresist structures. In order to anneal the ceramic microstructures without damaging, the dissolving behaviour of standard novolak photoresist was empirically assessed for different solvents. As a result, the resist was dissolved in tetrahydrofurane before pyrolysis. The processing of ceramic powders in the form of suspensions as used for bulk ceramic parts is new to the field of microfabrication.

5.1. Introduction

Ceramic materials in electronics and micro-electromechanical systems(MEMS) find extensive use as insulating layers, coatings or as dielectric layersin capacitor designs for dynamic random access memories (DRAM).[1] Prac-tically all of these are deposited by various thin film techniques such as chem-ical and physical vapour deposition, sol-gel or other coating technologies.The use of ceramic powders for generating microstructured ceramic elementsis new in the field of MEMS devices. For instance, micromachined gas sen-sors with a tin oxide thick-film layer deposited as colloidal suspension havebeen reported.[2,3] The deposition methods predominantly in use with thickfilm ceramics are different drop coating methods and screen printing. It isvirtually impossible to control the geometry of drop coated layers on the lowmicrometer level, whereas the practical resolution limits of screen printing lie

5.

102 C H A P T E R 5

in the range of 0.1 mm. We have recently introduced novel microfabricationtechniques using standard powder-based ceramic suspensions for generatingfeatures in the low micrometer range.[4-6] Promising results were obtainedalso by using soft lithography.[7,8] In this paper we report on a practicalapplication based on the filling of photoresist structures by casting of colloi-dal suspensions. Except for standard equipment to perform photolithogra-phy, the methods presented neither require vacuum technology nor otherexpensive devices and can be carried out in a general laboratory. Ceramicmicrolines of tin oxide with a cross sectional area of approx. 5 by 10 µm2

were produced. The dissolution behaviour of the novolak/diazoquinone pho-toresist material has been characterized empirically also for solvents not com-mon in standard photolithography processing steps.

5.2. Experimental

Photoresist structures for casting: a thick film novolak/diazochinone basedpositive photoresist (AZ4562, Clariant GmbH, Wiesbaden, Germany) wasspun at 3500 rpm (thickness of 6 µm) onto silicon test wafers. Prebake wasperformed at 100°C for 30 min. Using a Suss MA-6 aligner, line patterns fea-turing line widths of 10 m were transferred by UV illumination through achrome mask and developed with Microposit 351 basic solution (ShipleyInc., Marlborough, MA, USA). Before casting the ceramic microstructures,the open silicon surfaces of the pits were surface-activated by 2 min of oxygenplasma treatment.

Tin oxide suspension: pure tin oxide powder with a grain size ofd50 = 277 nm as determined by a Microtrac UPA 150 particle sizer was usedas received (Cerac Inc., Milwaukee, WI, USA). Polyacrylic acid sodium salt5100 (1.84 g, Fluka AG, Buchs, Switzerland) was dissolved in water (6.65 g,Millipore 18 MΩ cm) and 0.06 g ammonia (25% in H2O) were added, ris-ing the pH from 7.4 to 9.9. The solution was filtered through a 200 nm poresize teflon filter and subsequently used as additive. Tin oxide powder (21.1 g)was then slowly dispersed in 6.01 g of water containing 150 ml of additivesolution. The admixture was frequently interrupted by ball milling stepsusing zirconia balls.

P H O T O R E S I S T C A S T I N G 103

Dissolution tests. All solvents were of analytical grade and used as received.Some samples from photolithography were used without further treatment,others were postbaked at 120°C/30 min resp. 150°C/30 min (see Table 5–1).Dissolution times were assessed by immersing wafer pieces of approx.5 by 8 mm2 with developed and well visible resist structures in 12 ml of sol-vent. Time was recorded after the observation of complete disappearance ofall structural features. After 24 hrs, the undissolved resist samples were driedfor 12 hrs at 60°C and inspected.

Fig. 5–1. Schematic of the ceramic microfabrication process. a) colloidal suspension of tin oxide is cast into the pits of a 6 m thick struc-tured photoresist mask (b) using a soft polymer blade. c) remainders are wiped by a wetted blade. d) mask is dissolved, leaving only a thin resist layer after drying which fixes the ceramic green body. e) all resist material is pyrolysed and the ceramic microstructures are sintered.

e

da

c

b

104 C H A P T E R 5

5.3. Results and Discussion

The fabrication process is schematically summarized in Fig. 5–1. Crucial tothe whole process was the ability to induce different wetting properties of thematerials involved. Since the tin oxide suspension is based on water, its con-tact angle on the photoresist layer was determined to be 89°. Using a cut elas-tomeric blade of poly-dimethylsiloxane (contact angle to water 120°), it waspossible to drag a droplet of suspension over the substrate without leavingsignificant powder traces. The recessed line patterns could be filled com-pletely independent of the moving direction of the blade. However, somepowder remainders around the pits were always present on a small scale, seeFig. 5–2 In order to remove them, the blade was wetted with water vapourand repeatedly drawn over the substrate until the resist surface was clean.

Table 5–1. Empirically observed dissolution behaviour of 6 µm thick AZ4562 photoresist structures upon immersion in commonly used organic solvents. Prebake was performed at 100°C/ 30 min for all samples, some were postbaked at 120°C/ 30 min or 150°C/ 30 min respectively.

Solvent Polarity Observed Dissolving Times /sec

ETN a

a. Solvents are ordered according to polarity as determined by the solvatochromic effect on the π-π* transition energy observed in UV/VIS spectroscopy of pyridinium-N-phe-noxide betaine dye.[11] Normalised values are given.

100°C prebake

120°C postbake

150°C postbake

Water 1.000 - - -

Acetone 0.355 0-3 0-3 0-3

Ethyl acetate 0.228 0-3 10-15 10-15

THF 0.207 10-15 10-15 10-15

Toluene 0.099 -b

b. Denotes: insoluble during 24 hrs.

- -

n-Hexane 0.009 - - -

P H O T O R E S I S T C A S T I N G 105

Another important issue was the ceramic processing. In classic ceramics man-ufacturing, a ceramic suspension is dried after casting into shape, forming aso-called green body.[9] Green bodies consist of highly porous, non-consoli-dated network of powder and are therefore fragile. Higher density and char-acteristic ceramic properties are then reached by sintering at hightemperatures. Polymeric binders usually are added to enhance the mechanicalstrength of green bodies. In our case the polyacrylic acid served as a weakbinding agent in the dry state. In earlier thermogravimetry experiments itwas found that AZ4562 photoresist pyrolyses between 400°C and 500°C.However, the microstructures could not be sintered directly because of ther-mal expansion of the resist. Before pyrolysis it exerted too much pressure onthe green microstructures and destroyed them. In order to find a suitable sol-vent that dissolves the resist without damaging the fragile ceramics by liberat-ing tin oxide particles, a series of empirical dissolution tests was run. Theresults are shown in Table 5–1. Only solvents of intermediate polarity areable to dissolve AZ4562 quickly. Highly polar solvents as water and apolarsolvents as toluene and hexane fail. By postbaking the samples, i.e. cross-link-ing the novolak, a reduction in solubility could be expected. For the postbaketemperatures investigated, the effect was found to be of minimal influence.All these findings are consistent with the chemical structure of Novolak/diaz-oquinone resists.[10] For the third step (Fig. 5–1) therefore, tetrahydrofuranewas chosen, the solvent with the largest polarity difference to water that stilldissolves the resist.

Fig. 5–2. Microscope image of filled line-shaped pits of photoresist (dark, witdth 10 µm) before removal of powder traces atop.

106 C H A P T E R 5

After gently dipping the substrate filled with ceramic green body into tet-rahydrofurane for 30 s, the sample was allowed to dry. A very thin resist coat-ing remained, covering the whole surface including the ceramics. This layerthen provided some additional fixation during the early stages of the pyroly-sis/sintering process which was carried out by heating slowly to pyrolysis tem-perature (500°C) followed by raising the temperature to the sintering step(800°C/5 hrs). Subsequently, the resulting ceramic microlines were cleaned ofloose powder remains by rinsing with water and treating in a weak ultrasonicbath.

SEM images of the resulting structures are shown in Fig. 5–3. Thesemicrostructures exhibit significant differences to structured thin film ceram-ics: First, the aspect ratio lies in the range of one without the disadvantage oflong deposition times in thin film technology. Secondly, such powder-basedceramic microstructures may retain a certain degree of porosity and thus ahigh specific surface area, a feature previously unfamiliar to MEMS sytems.Thirdly, the powder synthesis and processing are separated. Therefore, a pow-der with well defined micro-, nano- and crystal structure or exactly tunedstoichiometry may be prepared and then used in microfabrication.

Fig. 5–3. SEM images of sintered ceramic microlines. a) view from top is given. b) close-up view of a line end seen from a tilting angle of 45°. The three-dimensional shape of the line, which is approx. 5 µm high and 10 µm in width, is evident.

P H O T O R E S I S T C A S T I N G 107

5.4. Summary

Ceramic microstructures of 10 µm size were successfully fabricated usingcasting of tin oxide powder suspension into photolithographically definednovolak/ diazochinone resist structures. The use of ceramic powders and theabundant experience available in powder synthesis probably will provideMEMS systems with new capabilities that are based on functional ceramics,with catalytic activity, controlled porosity, pyroelectricity, biocompatibilityand others. Future experiments will focus on compatibility issues to othermicrofabrication methods and MEMS systems.

108 C H A P T E R 5

5.5. References

[1] O. Auciello, R. Ramesh, Electroceramic Thin Films Part II: Device Applications, MRS Bulletin, 1996, 21, 7, 29.

[2] D. Briand, A. Krauss, B. van der Schoot, U. Weimar, N. Barsan, W. Gopel, N. F. de Rooij, Design and fabrication of high-temperature micro-hotplates for drop-coated gas sensors, Sens. Actuators B, 2000, 68, 1-3, 223-233.

[3] H. L. Tuller, Advanced Sensor Technology Based on Tin Oxide Thin Film - MEMS Integration, J. Electroceram., 2000, 4, 2, 415-425.

[4] U. Schönholzer, R. Hummel, L. J. Gauckler, Microfabrication of Ceramics by Fill-ing of Photoresist Molds, Adv. Mater., 2000, 12, 17, 1261-1263.

[5] U. Schönholzer, L. J. Gauckler, Ceramic Parts Patterned in the Micrometer Range, Adv. Mater., 1999, 11, 8, 630-632.

[6] U. Schönholzer, "Microfabrication of Ceramics", PhD thesis, 2000, ETH, Zurich.[7] Y. Xia, G. M. Whitesides, Soft Lithography, Angew. Chem. Int. Ed., 1998, 37, 550-

575.[8] M. Heule, L. Meier, L. J. Gauckler, Micropatterning of Ceramics on Substrates

towards Gas Sensing Applications, Mat. Res. Soc. Symp. Proc., 2001, Vol. 657, EE9.4.

[9] J. S. Reed, "Principles of Ceramics Processing", 2nd ed., Wiley, New York, 1995, 33.

[10] A. Reiser, H. Y. Shih, T. F. Yeh, J. P. Huang, Novolak-Diazoquinone Resists: The Imaging Systems of the Computer Chip, Angew. Chem. Int. Ed., 1996, 35, 2428-2440.

[11] C. Reichardt, "Solvents and Solvent Effects in Organic Chemistry", VCH, Wein-heim, 1988, 359.

109

M. Heule, K. Rezwan, L. Cavalli, L. J. Gauckler,Adv. Mater., 2002, submitted.

Miniaturised Enzyme Reactor Based on Hierarchically Shaped Porous Ceramic Microstruts

A microfluidic continuous flow enzyme reactor consisting of aluminum oxide microstruts mounted inside polydimethylsiloxane (PDMS) channels was built. Horse radish peroxidase was immobilised covalently on the alu-mina. Microstruts inside the fluidic channel were fabricated by casting a suspension of alumina powders into a developed photoresist mask. By this simple process, a two-fold hierarchical structure was obtained. First, the struts enforce a mixing by multiple splitting of the laminar flow and sub-divided the main PDMS channel into many 5-15 µm wide channels. Sec-ond, the struts consist of loosely sintered particle networks exhibiting well defined pores of 60 nm and a pore volume fraction of 23%. Compared to a microchannel system without these struts, a five-fold increase in enzy-matic product formation was obtained using off-line fluorescence spec-trometry.

6.1. Introduction

Heterogeneous enzymatic reactions such as ELISA-type immunoassays haverecently been performed in microfluidic devices.[1,2] These miniaturised ver-sions benefit from the larger surface-to-volume ratio of microchannel geome-tries and cause much lower cost due to the small amounts of reagentsnecessary. At the same time, the separation of enzyme and product after theincubation step is ensured by immobilisation of the enzyme, respectively the

6.

110 C H A P T E R 6

antibodies. Additional procedure steps like reagent preparation and detectionof products have also been integrated to form so-called “labs-on-a-chip” or“micro-total analysis systems” (µ-TAS).[3,4]

However, since liquid flow inside small channels is strictly laminar underalmost any practical conditions,[5] the mixing of reactands, respectively theaccess of substrate molecules to the enzyme becomes more difficult. The mix-ing is diffusion-limited in the absence of turbulent flow. Although it is possi-ble to elongate the channel system in order to wait for diffusion to completeas the fluid is transported through the microsystem, there have been severalstrategies presented to speed up mixing processes actively in microfluidic sys-tems. Bessoth et al. split a liquid flow into even smaller channels, where thediffusion distances become so small that the corresponding mixing times bydiffusion are in the order of milliseconds. Later, all divided flows are rejoinedagain.[6] Stroock et al. suggest to induce chaotic perturbations in the flowusing staggered herringbone structures at the channel bottom.[7] Also, activemixing schemes like ultrasonic agitation have been proposed.[8] A very prom-ising design has been put forward by Rohr et al. who created structures ofporous monolithic polymers by in-situ photo-initiated polymerisation of 2-hydroxymethyl methacrylate and ethylene dimethacrylate.[9]

In this article, we present the fabrication and initial testing of a micro-reactor design for heterogeneous reactions consisting of porous, accuratelyplaced microstruts made of ceramic aluminum oxide powders onto whichhorseradish peroxidase was immobilised. The flow channels were defined inpoly-dimethyl siloxane (PDMS) as commonly used in microfluidics and softlithography.[10]

M I N I A T U R I Z E D E N Z Y M E R E A C T O R 111

Fig. 6–1. a) schematic of the micro-strut layout. The microfluidic channel was 2000 µm wide and 15 µm deep, segmented by 9000 porous ceramic microstruts. The struts are organised in rows which are staggered to split the flow at every passage of rows. b) schematic cross-section of the micro-enzymereactor setup.

With the proposed design, we combine advantages of the approaches dis-cussed above. The porosity of the loosely sintered ceramic framework ensuresa large specific surface area and therefore a higher loading factor of enzymeper volume unit. The gaps between struts create multiple channels of 15-5 µm width. These distances can be covered by diffusion within short times(ms). All struts are organised in rows which are staggered, thus splitting the

Inlet

Outlet

10mm

Laminar flow split atevery row crossing

Single micro-strut: 15 x 200µm2

a

b

acrylic glass

acrylic glass

PDMSaluminamicrostruts

sapphire substrate

112 C H A P T E R 6

laminar flow and thus exposing the central region of the stream to theceramic enzyme-coated surface. Additionally, the ceramic parts of the devicecan be recycled by simply pyrolysing all protein and organic coatings at 800-900°C. A schematic including various blow-ups of the design is given inFig. 6–1. The conversion performance of the reactor containing microcer-amic elements is compared with an identical “empty” microchannel setup ofwhich only the bottom area was coated with immobilised enzyme.

It is already well established that covalent immobilisation of enzymes canbe performed on ceramic surfaces without substantial loss of enzyme activityby e.g. using the glutaraldehyde spacer method.[11,12] These ceramic micro-structures may provide interesting tools for various research applications. Forexample in implant materials research, it has been shown that osteoblastgrowth is even stimulated by titanium oxide ceramics[13,14] and influences ofsurface topography in the 10-50 µm range was characterized as well.[15,16]

6.2. Experimental

Photoresist Casting (PRC). Sapphire single crystal wafers (diameter 29 mm,thickness 0.8 mm) were spin-coated twice with AZ4562 thick-film positivephotoresist (Clariant GmbH, Wiesbaden, Germany) resulting in a total resistlayer thickness of 16-17 µm. The photomasks for the ceramic micro-ele-ments and the channel layout were designed using Adobe Illustrator (tm). Toobtain masks for UV exposure, the designs were printed with 5000 dpi reso-lution (ca. 5 µm dot size) onto offset printing transparencies by a professionalprinting service. The transparencies were cut and mounted onto 5 inchsquare glass plates compatible with a Suss MA-6 mask aligner (Suss GmbH,Munich, Germany). After exposure and development, a 50 vol% suspensionof aluminum oxide was cast into the recessed photoresist structures. For thesuspension, 192 g Al2O3 Taimicron TM-DAR 2831, mean particle sized50 = 190 nm, as obtained from Taimei Chemicals, Tokyo, Japan was dis-persed in 46 ml H2O containing 1.1 g HCl 5 M and 0.9 g methylated-cellu-lose. After drying for 60 min, the photoresist was dissolved by dipping thewhole substrate for 7 s into tetrahydrofurane. Finally, the structures were sin-tered by heating (heating rate 2.5°min-1) to 1000°C and holding at that tem-perature for 120 min.

M I N I A T U R I Z E D E N Z Y M E R E A C T O R 113

Flow channel definition. The microfluidic channel was fabricated by castingpoly-dimethylsiloxane (PDMS, Sylgard 184, Dow Corning) against a masterstructure of exposed and developed photoresist, which was obtained usingthe same transparency mask technology as described above.

Enzyme immobilisation. All chemicals were obtained from Fluka AG,Buchs, Switzerland and used without further treatment. The microstructuredsapphire substrates were first immersed in 3-aminopropyltriethoxysilane(0.43 M in H2O set to pH = 3.5 by adding HCl) for 60 min and heated to115°C for 90 min. Then, the samples were immersed for 60 min in glutaral-dehyde (hexane-1,6-dialdehyde, 2.6 M) containing phosphate buffer solution(PBS, 3.55 g Na2HPO4 in 500 g H2O, 0.05 M, set to pH = 7.0 by addingHCl). Finally, the enzyme coupling was carried out by immersing the sub-strates for 300 min in a solution prepared with 3.4 mg horseradish peroxidase(588 U mg-1) and 45 ml PBS.

Microreactor assembly. For the liquid inlet and outlet, holes were cutthrough the PDMS blocks containing the recessed flow channels on the bot-tom. The PDMS channels were aligned with respect to the ceramic micro-structures using binoculars and a xyz-rotation stage, then brought intocontact and secured by screwing an acrylic glass plate with plugs for PE tub-ing on top.

Enzymatic reaction on the chip. Using a syringe pump, a substrate solu-tion of 1.5 mM homovanillic acid (HVA) and 0.5 mM H2O2 in PBS bufferwas pumped with a flow rate of 160 µl hr-1. The product-containing solutionwas gathered in a PE tube after streaming through the microreactor. Every30 min, 80 µl of product solution were removed and injected into a 96 wellplate. Before analysis in an automated luminescence spectrometer (PerkinElmer LS 55 in fluorescence mode, excitation wavelength 310 nm, readingwavelength 430 nm, slit width 2.5 nm), 120 µl of PBS were added to bringthe volume in each well to a total of 200 µl (to avoid fluorescence quenchingand other volume effects).

Porosimetry. Remainders of the alumina suspension were cast and sinteredunder identical conditions as stated above. Fragments of 2-3 mm size wereanalysed in a mercury intrusion type porosimeter (Carlo Erba Instruments2000 Porosimeter).

114 C H A P T E R 6

6.3. Results and Discussion

For this project, we tried to use simple microfabrication methods. A fabrica-tion scheme we refer to as photo resist casting (PRC) was applied in order toreach the smallest feature size in the range of 10-20 µm. As we have shownpreviously, ceramic microstructures of 10 µm size with an accuracy of 1-2 µmcan be fabricated.[17,18] Ceramic suspensions with high solids loading as usedfor traditional ceramic slip casting processes are cast into the recessed struc-tures of a developed photoresist mask. Therefore, microstructures of the verysame ceramic material as e.g. an implant part may be produced. For thismodel application, we used a standard aluminum oxide suspension of50 vol% solids loading, which is a lower value than usual. It was decided touse 50 vol% in order to create a porous rather than a dense ceramic network.

Fig. 6–2. SEM oblique views on the microceramic elements. a) overview onto a large section of the channel after sintering and before enzym immobilisation. b) close-up showing an undercut sidewall from the same angle as in a), revealing the 5 µm domes resulting from the transparency photomasks used. c) SEM showing the microstructure of the loosely sintered 200 nm particles. d) fully assembled microreactor through which ink-solution is pumped for visualising the channel layout.

M I N I A T U R I Z E D E N Z Y M E R E A C T O R 115

In Fig. 6–2a, an oblique SEM view on a section of the micro-strut coveredarea is given. The dotted line indicates where the PDMS channel boundarywould be placed when sealing the device. The photomask for UV-exposure ofthe pattern was a high-resolution printed transparency ordered from a profes-sional offset-printing service. A similar simple-photomask approach usingmicrofiche has been suggested by Deng.[19] That procedure reduced cost forone new design to only a few dollars and more significantly, a full develop-ment cycle from a new computer drawing to the final ceramic microstruc-tures could be completed within three days. The resolution on thetransparency was 5000 dpi, which corresponds to a single dot size of ca.5 µm. As observed in the light microscope, the printer rendered a smoothlydesigned line with an edge roughness of about the same 5 µm. Therefore, weestimate the lower practical limit of structure reproduction around 15-20 µm. It was also necessary to compensate black and white (unprinted)areas, since the printer’s black areas grew ca. 5 µm wider than originallydesigned. When the 17 µm thick photoresist layer was exposed through thetransparency mask, UV-scattering at structure egdes occurred. Due to a highUV-dose necessary to expose the whole photoresist thickness, an inclined androunded sidewall profile resulted upon development. Therefore, the resultingceramic microstruts obtained by filling the photoresist structures withceramic powders exhibit a large pagoda-roof shaped undercut as depicted inFig. 6–2a. Sapphire wafers (Al2O3 single crystal) were used to match the ther-mal expansion coefficient of the alumina powders. Cracks in the struts fromthe sintering process at 1000°C could thus be fully avoided. Due to its trans-parency, the channels could be observed in operation from below using aninverted microscope (Zeiss Axiovert 25). In principle, opaque but muchcheaper polished alumina platelets could be used as well. In Fig. 6–2c, a closeup of the porous structure after the sintering process is shown. The particlesize still remains in the 200 nm range as in the original Al2O3 powder. Occa-sionally, two or three grains grown together can be observed. This porousnetwork of particles provided a large specific surface for enzyme immobilisa-tion. The results from mercury intrusion porosimetry are shown in Fig. 6–3.There is a sharp increase in absorbed Hg volume at a pressure equivalent to apore radius of 29 nm. We therefore observed a well-defined pore diameter of

116 C H A P T E R 6

approx. 60 nm and a total volume of porosity of 23% which is accessible forenzyme and substrate. The method was also used to estimate the specific sur-face due to this open porosity to be in the range of 5.3 m2/g (the originalpowder had 14.3 m2/g). This would correspond to a larger surface by a factorof 60 owing to the porous ceramics.

Immobilisation of horseradish peroxidase was carried out only with thesapphire wafer by using well known chemistry. First, the Al2O3 surface wasfunctionalised with Amino-groups (using 3-aminopropyltriethoxysilane).Then, the bifunctional glutaraldehyde was added forming free aldehyde ends,which were immediately coupled to horseradish peroxidase. Finally, thePDMS microchannels were aligned to the microstructures and sealed bypressing the PDMS-sapphire sandwich between two acrylic glass plates (seeFig. 6–2d).

Fig. 6–3. Results from the mercury intrusion porosimetry. From recording the pressure-dependence of Hg intrusion into the ceramic pores, equivalent pore radii are calculated.

Peroxidases are robust enzymes which convert a reductant in conjunctionwith a peroxide according to the following general equation.[20]

(1)

µ

8

7

6

5

4

3

2

1

0

Pore

Vo

lum

e /

mm

3 g-1

0.012 3 4 5 6 7 8 9

0.1Pore Radius/ m

100

80

60

40

20

0

Pore V

ol. cu

mu

lative /%

ROOH 2AH+ ROH H2O A2+ +→

M I N I A T U R I Z E D E N Z Y M E R E A C T O R 117

For assessment of the overall enzymatic reaction performance, a reactionforming a fluorescent product from non-fluorescent reductants was used. Inthis case, the reductant was homovanillic acid (4-hydroxy-3-methoxy-pheny-lacetic acid, HVA) forming the fluorescent biphenylic product 2,2’-dihy-droxy-3,3’-dimethoxybiphenyl-5,5’-diacetic acid with hydrogen peroxideanalogous to Eq. (1), see also Fig. 6–4a for the schematic of the reaction.This system has also been tested before.[21] For assessing the enzymatic activ-ity, a solution containing HVA and H2O2 was pumped through the microre-actor at a flow rate of 160 µl min-1 by a standard syringe pump. In allexperiments, both versions were driven simultaneously: the reactor withceramic microstructures and a reference reactor consisting only of a plain sap-phire wafer onto which horseradish peroxidase was immobilised. The detec-tion of product was performed offline, i.e. the product solutions were firstcollected in a PE tube connected to the outlets and later transferred to amicro-well plate, whose 96 wells could be read automatically within 1 min bythe fluorescence spectrometer. Optimal excitation wavelength was deter-mined in preliminary experiments to be at 310 nm, the fluorescence intensitywas measured at 430 nm. Calibration series with known biphenyl concentra-tions showed an accuracy of 10% in the µM range. The bar chart in Fig. 6–4b shows the results obtained from four consecutive runs on the samemicroreactors. Between each run, the microchannels were flushed thoroughlywith buffer solution (PBS) before injecting new substrate solution. Startingwith the third run, a freshly prepared substrate solution was used (markedwith asterisk *). On the right hand chart, the fluorescent background intensi-ties of individual reactand species in the same concentration as in the sub-strate solution are given. No reactand caused a fluorescence signal above 100intensity units, the average being about 60. The first batch of 80 µl productsolution had an intensity of almost 500, whereas the product solution fromthe reference microreactor without the ceramic structures only reached 100, avalue only slightly above the background reading. Expressed in absolute con-centrations obtained from large-scale calibration experiments, this first batchwas 65 ± 10 µM in fluorescent product (yield 13% of 500 µM available incase of full conversion). This also corresponds to a biphenyl production

118 C H A P T E R 6

increased by a factor of 5 in the micro-strut reactor. Over the course of a fewhours in subsequent runs, the ratio was not as high, due to loss of enzymeactivity and probably some ageing in the substrate solution.

With photo resist casting we realised a new microfluidic reactor based onporous ceramic microstruts onto which enzymes were immobilisedcovalently. The design of the hierarchical arrangement of the ceramic powderas a microstructure inherently takes into account problems associated withminiaturisation of heterogeneous reactions like the mixing of laminar flowsand offers a very large specific surface for the enzymatic reaction due to itsporosity. A first comparison to a channel system without these microelementsshowed a 300-500% increase in product formation of an enzymatic reaction.

Fig. 6–4. a) reaction of HVA to fluorescent product catalysed by horserad-ish peroxidase. b) fluorescence intensity results of different exper-iments with background signals for reference on the right. See text for details.

HVA H2O2 H2O HVA/H2O2

600

500

400

300

200

100

0

flu

ore

scen

ce in

ten

sity

/a.u

.

1st 2nd 3rd* 4th*

COOH

H2O+ 2

+ 2

horseradishperoxidase

COOH

MeO

OH

MeO

HO

HOOC

OMe

HO

H2O2

λex = 310nm, λdet = 430nm

a

bwith Al2O3 microstructures

empty channel

experiment run

reference:background fluorescence

M I N I A T U R I Z E D E N Z Y M E R E A C T O R 119

6.4. References

[1] H. Mao, T. Yang, P. S. Cremer, Design and Characterization of Immobilized Enzymes in Microfluidic Systems, Anal. Chem., 2002, 74, 2, 379-385.

[2] E. Eteshola, D. Leckband, Development and characterization of an ELISA assay in PDMS microfluidic channels, Sens. Actuators B, 2001, 72, 129-133.

[3] P. A. Auroux, D. Iossifidis, D. R. Reyes, A. Manz, Micro Total Analysis Systems. 1. Introduction, Theory and Technology, Anal. Chem., 2002, 74, 2623-2636.

[4] P. A. Auroux, D. Iossifidis, D. R. Reyes, A. Manz, Micro Total Analysis Systems. 2. Analytical Standard Operations and Applications, Anal. Chem., 2002, 74, 2637-2652.

[5] G. M. Whitesides, A. D. Stroock, Flexible Methods for Microfluidics, Physics Today, 2001, 54, 6, 42-48.

[6] F. K. Bessoth, A. J. de Mello, A. Manz, Microstructure for efficient continuous flow mixing, Anal. Commun., 1999, 36, 213-215.

[7] A. D. Stroock, S. K. W. Dertinger, A. Ajdari, I. Mezic, H. A. Stone, G. M. White-sides, Chaotic Mixer for Microchannels, Science, 2002, 295, 647-651.

[8] Z. Yang, S. Matsumoto, H. Goto, M. Matsumoto, R. Maeda, Ultrasonic micro-mixer for microfluidic systems, Sens. Actuators A, 2001, 93, 3, 266-272.

[9] T. Rohr, C. Yu, M. H. Davey, F. Svec, J. M. J. Frechet, Porous polymer monoliths: Simple and efficient mixers prepared by direct polymerization in the channels of microfluidic chips, Electrophor., 2001, 22, 18, 3959-3967.

[10] Y. Xia, G. M. Whitesides, Soft Lithography, Angew. Chem. Int. Ed., 1998, 37, 550-575.

[11] H. Weetall, Covalent Coupling Methods for Inorganic Support Materials, Methods in Enzymology, 1976, XLIV, 134-149.

[12] S. W. Park, Y. I. Kim, K. H. Chung, S.W. Kim, Improvement of stability of immo-bilized GL-7-ACA acylase through modification with glutaraldehyde, Process Bio-chemistry, 2001, 37, 153-163.

[13] B. Fartash, H. Liao, J. Li, N. Fouda, L. Hermansson, Long-Term Evaluation of Titania-Based Ceramics Compared With Commercially Pure Titanium in-Vivo, Journal of Materials Science-Materials in Medicine, 1995, 6, 8, 451-454.

[14] J. G. Li, Behavior of Titanium and Titania-Based Ceramics Invitro and Invivo, Biomat., 1993, 14, 3, 229-232.

[15] B. Chehroudi, D. McDonnell, D. M. Brunette, The effects of micromachined sur-faces on formation of bonelike tissue on subcutaneous implants as assessed by radi-ography and computer image processing., J Biomed Mater Res, 1997, 34, 3, 279-90.

[16] B. Chehroudi, D. M. Brunette In Encyclopedic Handbook of Biomaterials and Bioengineering. Part A; D. L. Wise, Ed.; Marcel Dekker Inc.: N.Y., 1995; Vol. 1.

[17] U. Schönholzer, R. Hummel, L. J. Gauckler, Microfabrication of Ceramics by Fill-ing of Photoresist Molds, Adv. Mater., 2000, 12, 17, 1261-1263.

[18] M. Heule, L. J. Gauckler, Microfabrication of Ceramics based on Colloidal Sus-pensions and Photoresist Masks, J. Photopolym. Sci. Technol., 2001, 14, 3, 449.

[19] T. Deng, J. Tien, B. Xu, G. M. Whitesides, Using Patterns in Microfiche as Photo-masks in 10 µm-scale Microfabrication, Langmuir, 1999, 15, 6575-6581.

120 C H A P T E R 6

[20] L. Stryer, "Biochemistry", fourth ed., Freeman, New York, 1995, p. 553.[21] F. Wang, F. Schubert, H. Rinneberg, A fluorometric rate assay of hydrogen perox-

ide using immobilized peroxidase with a fibre-optic detector, Sens. Actuators B, 1995, 28, 3-7.

121

M. Heule, L. J. Gauckler, Adv. Mater.,2001, 13 (23), 1790 -1793.

Science, Editors’ Choice, „Tiny Gas Sensors,”2002, 295, 15.

Gas Sensors Fabricated from Ceramic Suspensions by Micromolding in Capillaries

Micromolding in capillaries was used to fabricate an array of SnO2 gas sensors using colloidal suspensions. The micromolding procedure was inte-grated in a sequence of conventional photolithography steps. Miniaturized gas sensors on sapphire single crystal substrates showed responses for 100 ppm of hydrogen and 600 ppm for carbon monoxide obtained from an active sensing area of only 10 by 40 µm2.

7.1. Introduction

Extending the range of available materials has become an integral part in thedevelopment process of new Microelectromechanical Systems (MEMS). Inrecent years, research efforts have been devoted to the development ofadvanced microsystems that are able to perform optical, biological or chemi-cal functions.[1-3] Apart from the traditional semiconductor materials, metaloxide powders offer many opportunities to enhance the functionality ofMEMS. Miniaturization of solid-state chemical gas sensors based on semi-conducting oxides has reduced power consumption to a degree that hasallowed the employment of the gas sensors in miniaturized analytical devices.Today, the active ceramic layer is usually prepared using thin film depositiontechniques, screen printing or droplet deposition processes like ink jet print-ing.[4-6] Tin oxide is the best-characterized material for solid-state gas sensors

7.

122 C H A P T E R 7

that have been in use for many years already.[7,8] Changes in electrical resis-tance are measured when a combustible analyte gas reacts catalytically withoxygen species of the ceramic surface.[9]

Soft lithography has provided a valuable tool for the micro-structuring ofliquid materials at low cost.[10-12] Patterned ceramic materials on surfaceshave been created by soft lithographic techniques like Micromolding in Cap-illaries (MIMIC)[13] or micro-contact printing, using sol-gel or polymericprecursor processing routes. Sol-gel patterned films of Sr2Nb2O7 were pro-cessed with MIMIC[14] and polymeric precursors for SiBNC ceramics.[15]

SnO2 and ZrO2 microstructures derived from MIMIC were reportedalso.[16] Immobilized zeolithes in well-defined surface patterns on siliconsubstrates were fabricated using micro- contact printing.[17] Recently we havegenerated ceramic microstructures from colloidal suspensions featuringdimensions in the order of 10 µm.[18,19] Some examples of structures andpatterns have also been reported for functional devices.[20,21]

In the present work we demonstrate the application of MIMIC for a sim-ple thick-film gas-sensing device based on tin oxide by combining photoli-thography and MIMIC with ceramic suspensions. We report on gas sensingelements covering an area of only 10 by 40 µm2, which includes contacts byplatinum electrodes to perform resistance measurements. This is a factor of50 to 100 smaller than in use today.

7.2. Experimental

Substrates: Sapphire single crystal wafers of 29 mm diameter were rinsedwith water (18 MΩ cm, Millipore), cleaned in piranha acid solution (7 partsconcentrated sulfuric acid, 3 parts hydrogen peroxide 35%) at room temper-ature for 12 hrs, followed by water.

Photolithography: In all three photolithography steps (heater, wires, oxideetch) the following procedures were applied: AZ4562 positive thick film pho-toresist (Clariant GmbH, Wiesbaden, Germany) was spun at 3500 rpm for40 s and pre-baked at 100°C / 30 min, leading to a thickness of approx.6 µm. UV illumination (dose 700 mJ cm-2) for pattern transfer was per-

T I N Y G A S S E N S O R S 123

formed using a Suss MA-6 aligner (Suss GmbH, Munich, Germany) anddeveloped in a 1:3 Microposit 351 - water mixture (Shipley Inc., Marlbor-ough, MA, USA).

Platinum deposition for the heater and electrode layers: Platinum wasevaporated using a BAE 370 e-beam evaporation system at 5·10-6 mbar(Unaxis, Balzers, Liechtenstein). Lift-off was performed by ultrasonication ofthe sample in acetone. The 200 nm insulating layer of silicon was also depos-ited with the BAE 370.

Micromolding in capillaries (MIMIC): 9 g of Sylgard 184 PDMS pre-polymer/catalyst mixture (Dow Corning Inc., Midland, MI, USA) waspoured over a photoresist structure and cured for 26 hrs. The PDMS moldwas oxygen plasma-treated for 2 min before use. The mold was then placedon the center over the heating structure. 25 µl of tin oxide suspension weredispensed at the entrance of the structures. After the capillary filling, thesample was allowed to dry for 30 min at room temperature. The dried sus-pension droplet at the entrance was then removed by pressing the filledPDMS mold tightly to the wafer and rinsing the outside region with plentyof water, immediately followed by acetone to quickly remove the slowly dry-ing water. After another 30 min, the PDMS mold was lifted and the samplewas fired at 800°C for 5 hrs.

Post-MIMIC processing: The contacting platinum wires were prepared byphotolithography performed on the previously deposited ceramic structures,followed by Pt deposition and lift-off. In the final step, AZ4562 resist wasused as an etch mask for silicon oxide etching in NH4-buffered hydrofluoricacid (VLSI Selectipur 12.5%, Merck KGaA, Darmstadt, Germany). Electri-cal contacts were set after mounting on a printed circuit board (PCB) bywedge-wire-bonding using 30 µm diameter aluminium wire on a semi-auto-matic wire bonding system Delvotec 5425 (F&K Delvotec GmbH,Ottobrunn, Germany).

Gas sensitivity measurements: Digital mass flow controllers (El-Flow Dig-ital, Bronkhorst Hitec, Ruurlo, Netherlands) were used for the mixing ratiosand to control the total gas flow. Filtered air containing 40% relative humid-ity was used as a reference atmosphere. The analyte gases were purchased ascertified mixtures of 2000 ppm in high purity nitrogen (PanGas AG, Dag-

124 C H A P T E R 7

mersellen, Switzerland). Depending on the desired dilution ratio, the totalgas flow rate was set from 150 ml min-1 to 316 ml min-1. In most cases theflow rates were fixed at 200 ml min-1. The sensor was heated to 280°C. Theelectrical resistance of the tin oxide micro-sensor was monitored using a con-ventional digital multimeter.

7.3. Results and Discussion

Fig. 7–1. Outline of the micro thick-film gas sensor fabrication process.

Sapphire single crystal plates of 30 mm diameter and a thickness of 0.9 mmwere used as a substrate. As schematically illustrated in Fig. 7–1, the overallprocess required three photolithography steps and the micro-structuring pro-cedure of tin oxide ceramics by Micromolding in Capillaries. A meander-shaped heating wire pattern was defined by conventional photolithography.After evaporation of a 120 nm platinum layer and then a lift-off procedure,the heating pattern was covered by a thin silicon oxide insulating layer of200 nm thickness (Fig. 7–1a). The tin oxide colloidal suspension containing33 vol% solids loading was optimized for low viscosity of 25 mPa s at ashearing frequency of 100 Hz. An elastomeric mold of poly-dimethylsiloxane

a) SiO2 insulating layer covering microstructured heater wire (Pt)

d) deposit Pt contact electrodes

f) packaging, wire bondingc) anneal at 800˚C / 5 hrs

e) SiO2 etch: open heater contactsb) Capillary filling of microlines.

PDMS

T I N Y G A S S E N S O R S 125

(PDMS) containing recessed microchannel lines was obtained by curingPDMS pre-polymer over a master structure of photoresist. When placed onthe sapphire single crystal substrate, the surface structures of the PDMS moldformed fine line-shaped capillaries. The inner surfaces of the capillaries wereoxygen plasma-treated to create a highly hydrophilic surface. These lines werefilled with ceramic suspension by capillary force action (Fig. 7–1b). Afterdrying the mold was removed, releasing green body microstructures that wereslightly sintered at 800°C for 300 min. Further details of this process havebeen published elsewhere.[22] Porous ceramic tin oxide lines resulted with awidth of 10 µm, height 5 µm and length 1 mm. Good adhesion of theceramic microstructures to the underlying silicon oxide layer made it possibleto clean the samples in water by ultrasound (Fig. 7–1c). Electrode patterns ofplatinum were photolithographically defined in the same manner as theunderlying heating structure. The substrate was coated with a 200 nm plati-num layer, followed by a lift-off process to form the contacts (Fig. 7–1d).Finally, the heater layer was accessed by etching the silicon oxide layer inbuffered HF solution (Fig. 7–1e). The sensor and heater were contacted byaluminium wedge wire bonding to the outside (Fig. 7–1f).

126 C H A P T E R 7

Fig. 7–2. a) top view of the gas sensor set-up seen through the photoresist mask before deposition of the Pt thin film wires for resistance measurement. Each substrate contained one heating element onto which ten connected ceramic microclines were applied (dark horizontal lines). b) isometric drawing representing the contact layout for one tin oxide microline. Up to four contacts per micro-line were available. The inner contacts labelled '1' were connected for the sensor response measurements presented in this article. c) SEM close-up view of micro thick-film gas sensors ready for measurement (uncoated sample, taken with 0.3 kV acceleration voltage on a LEO 1530 SEM, LEO Electron Microscopy Ltd, Cam-bridge, UK).

The specific order of deposition steps was chosen because our approach relieson sintering the ceramic powder. Due to high surface mobility of Pt at thesintering temperatures employed for the SnO2, the thin platinum contactswill form pinholes in air and finally degrade into small disconnectedislands.[23] However, covering the Pt-lines intermediately with a thin siliconoxide layer effectively protected the Pt-lines at 800°C. The individual tinoxide powder grains form a stable porous network (55% solid), which wasprerequisite to avoid contamination of the clean room environment with tin

T I N Y G A S S E N S O R S 127

oxide particles (steps d-f in Fig. 7–1). A 6 µm thick photo resist layer wasspin coated for defining the electrode wire layout. It also served as a protect-ing layer to prevent any loss of powder during handling of the substrates. Thesample was brought into contact with the photomask using minimal pressurefor UV-illumination (soft contact mode). The development process of theexposed substrate was performed as usual. A microscopy image of the result-ing pattern is given in Fig. 7–2a. The electrode design consisted of four indi-vidual connects per ceramic microline. Two connects entered on the long axisof the microline, stepping onto the line and running all the way on top to thecenter. Two other connects were designed to run parallel to the ceramicmicroline at 20 µm distance up to the center. At that point they bend by 90°and step onto the ceramic line covering an area of approximately10 by 10 µm2 on the SnO2 (Fig. 7–2b). Between each connect a gap of20 µm on the tin oxide microstructure was left uncoated to form the actualgas sensing area. A SEM close-up view of a micro-thick-film sensor ready formeasurement is given in Fig. 7–2c.

Thermal calibration of the heater was carried out using a few mg of sev-eral pure organic substances with well-defined melting temperatures (p-anisicacid 183°C, caffeine 235°C, anthrachinone 283°C). Due to the large heatdissipation into the sapphire single crystal substrate, a heating power as highas 2.2 W was required to reach a temperature of 280°C at the silicon oxidesurface. Drastic reduction in power consumption will be possible using amicro-hot plate design. Measurable damage to the sensor could be caused byapplying an electrical power in excess of 2.5 W in the present design.

For assessment of the gas sensitivity, the sensor was mounted in a gaschamber with a volume of 11 ml. The times needed to exchange the wholegas volume above the sensor were comparable to the response times of themass flow controllers upon changing the flow rate. The resistance betweentwo center electrodes of one ceramic microline was measured. Several dilu-tion series with analyte gas pulses of 2 min duration ranging from 100 ppmto 2000 ppm were run using hydrogen, carbon monoxide and methane.Fig. 7–3 shows the recorded sensitivities for hydrogen (Fig. 7–3a) and carbonmonoxide (Fig. 7–3b). The sensitivity is defined as ratio Rgas / Rair of themeasured resistance during gas exposition to the base resistance measured in a

128 C H A P T E R 7

reference atmosphere (air of 40% relative humidity in this case). A value ofRair = 8.3 · 106 Ω was recorded at T = 280°C. Correspondingly, with anatmosphere containing 2000 ppm of hydrogen RH2 = 2.6 · 106 Ω wasobtained. The response times were in the range of a few seconds and therecovery times

Fig. 7–3. a) sensor response at 280°C for hydrogen as function of time. The dashed bars indicate periods during which analyte gas was injected, showing the preset gas concentrations on the right-hand axis. b) sensor response for carbon monoxide.

1.0

0.8

0.6

0.4

0.2

0.0

R /

R0

500040003000200010000time /s

0

600

1000

2000

CO

/pp

m

1.0

0.8

0.6

0.4

0.2

0.0

R /

R0

40003000200010000time /s

0

600

1000

2000

H2 /p

pm

a

b

T I N Y G A S S E N S O R S 129

of the sensor signal upon changing the analyte-containing atmosphere backto air were approximately 8 min, as expected for tin oxide gas sensors. Thelimit of detection (LOD) is defined by the signal to noise ratio (s/n) of theacquired baseline data. A signal larger than 3 times s/n is usually required forquantitative detection of an analyte.[24] Hydrogen could be unambiguouslydetected at a concentration of 100 ppm, and CO at 600 ppm and above. Theconcentration vs. peak height relation in the range from 100 to 2000 ppmappears to roughly follow a linear trend for both analyte gases. The tin oxidepowder lines in the present study had not been doped with any transitionmetal except the part of the Pt-line contacts deposited directly onto the tinoxide surface that served as a catalyst. Their effect is clearly distinguishableregarding the different sensitivities obtained for hydrogen, which is stronglyactivated by platinum, compared to carbon monoxide. During the wholemeasurement, the baseline was observed to drift towards higher resistances by15%. We attribute this behavior to cooling of the tin oxide in the gas stream.

7.4. Summary

By combining soft lithography with finely tuned colloidal suspensions ofceramic powders and MIMIC we have demonstrated a miniaturized ceramicthick-film gas sensor based on tin oxide. Compared to other existing designs,the sensing area of 10 by 40 µm2 is 2 to 5 orders of magnitude smaller than atypical thick-film design prepared by standard thick-film deposition methodslike screen printing or droplet deposition[25] (200 · 200 µm2). Hydrogenand carbon monoxide could be detected in the range of 100 ppm. We dem-onstrated a method to combine photolithography and gas phase depositionwith Micromolding in Capillaries (MIMIC) to form the ceramic microstruc-tures. A complete sensor array was demonstrated to cover approximately thesame area as one gas sensor prepared by conventional micromachining. Thedemonstrated method is suitable for processing a variety of different powders.In principle, the fabrication rules established in traditional ceramic manufac-turing are still valid in the low micrometer regime. Properties of advancedfunctional ceramic materials like catalytic activity, biocompatibility or welldefined porosity may be integrated into MEMS devices not only in the formof coatings but also as micropatterned structures from powder suspensions.

130 C H A P T E R 7

7.5. References

[1] D. L. Dickensheets, G. S. Kino, Silicon Micromachined Scanning Confocal Opti-cal Microscope, IEEE J. MEMS, 1998, 7, 1, 38-47.

[2] J. Wang, M. P. Chatrathi, B. M. Tian, Micromachined separation chips with a pre-column reactor and end-column electrochemical detector, Anal. Chem., 2000, 72, 23, 5774-5778.

[3] C. H. Mastrangelo, M. A. Burns, D. T. Burke, Microfabricated Devices for Genetic Diagnostics, Proc. IEEE, 1998, 86, 8, 1769-1787.

[4] J. S. Suehle, R. E. Cavicchi, M. Gaitan, S. Semancik, Tin Oxide Gas Sensor Fabri-cated Using CMOS Micro-Hotplates and In-Situ Processing, IEEE Electron Device Letters, 1993, 14, 3, 118-120.

[5] G. Faglia, E. Comini, A. Cristalli, G. Sberveglieri, L. Dori, Very low power con-sumption micromachined CO sensors, Sens. Actuators B, 1999, 55, 140-146.

[6] D. Briand, A. Krauss, B. van der Schoot, U. Weimar, N. Barsan, W. Gopel, N. F. de Rooij, Design and fabrication of high-temperature micro-hotplates for drop-coated gas sensors, Sens. Actuators B, 2000, 68, 1-3, 223-233.

[7] N. Barsan, M. Schweizer-Berberich, W. Göpel, Fundamental and practical aspects in the design of nanoscaled SnO2 gas sensors: a status report, Fresenius J. Anal. Chem., 1999, 365, 287-304.

[8] K. Ihokura, J. Watson, "The Stannic Oxide Gas Sensor", CRC Press, New York, 1994.

[9] W. Göpel, O. Reinhardt, "Metal Oxide Gas Sensors: New Devices through Tailor-ing Interfaces on the Atomic Scale", edited by H. Baltes, J. Hesse, W. Göpel, Sen-sors Update, VCH, Weinheim, 1996, p. 49.

[10] Y. Xia, G. M. Whitesides, Soft Lithography, Angew. Chem. Int. Ed., 1998, 37, 550-575.

[11] W. S. Beh, I. T. Kim, D. Qin, Y. Xia, G. M. Whitesides, Formation of Patterned Microstructures of Conducting Polymers by Soft Lithography, and Applications in Microelectronic Device Fabrication, Adv. Mater., 1999, 11, 12, 1038.

[12] B. Xu, F. Arias, S. T. Brittain, X. M. Zhao, B. Grzybowski, S. Torquato, G. M. Whitesides, Making Negative Poisson's Ratio Microstructures by Soft Lithography, Adv. Mater., 1999, 11, 14, 1186.

[13] E. Kim, Y. Xia, G. M. Whitesides, Micromolding in Capillaries: Applications in Materials Science, J. Am. Chem. Soc., 1996, 118, 5722-5731.

[14] S. Seraji, Y. Wu, N. E. Jewell-Larson, M. J. Forbess, S. J. Limmer, T. P. Chou, G. Cao, Patterned Microstructure of Sol-Gel Derived Complex Oxides Using Soft Lithography, Adv. Mater., 2000, 12, 19, 1421.

[15] H. Yang, P. Deschatelets, S. T. Brittain, G. M. Whitesides, Fabrication of high per-formance ceramic microstructures from a polymeric precursor using soft lithogra-phy, Adv. Mater., 2001, 13, 1, 54.

[16] W. S. Beh, Y. Xia, D. Qin, Formation of patterned microstructures of polycrystal-line ceramics from precursor polymers using micromolding in capillaries, J. Mater. Res., 1999, 14, 10, 3995-4003.

[17] K. Ha, Y. J. Lee, D. Y. Jung, J. H. Lee, K. B. Yoon, Micropatterning of oriented zeolite monolayers on glass by covalent linkage, Adv. Mater., 2000, 12, 21, 1614.

T I N Y G A S S E N S O R S 131

[18] U. Schönholzer, L. J. Gauckler, Ceramic Parts Patterned in the Micrometer Range, Adv. Mater., 1999, 11, 8, 630-632.

[19] U. Schönholzer, R. Hummel, L. J. Gauckler, Microfabrication of Ceramics by Fill-ing of Photoresist Molds, Adv. Mater., 2000, 12, 17, 1261-1263.

[20] J. Hu, R. G. Beck, T. Deng, R. M. Westervelt, K. D. Maranowski, A. C. Gossard, G. M. Whitesides, Using soft lithography to fabricate GaAs/AlGaAs heterostruc-ture field effect transistors, Appl. Phys. Lett., 1997, 71, 14, 2020-2022.

[21] O. J. A. Schueller, X. M. Zhao, G. M. Whitesides, S. P. Smith, M. Prentiss, Fabri-cation of Liquid-Core Waveguides by Soft Lithography, Adv. Mater., 1999, 11, 37.

[22] M. Heule, L. Meier, L. J. Gauckler, Micropatterning of Ceramics on Substrates towards Gas Sensing Applications, Mat. Res. Soc. Symp. Proc., 2001, Vol. 657, EE9.4.

[23] S. L. Firebaugh, K. F. Jensen, M. A. Schmidt, Investigation of High-Temperature Degradation of Platinum Thin Films with an In Situ Resistance Measurement Apparatus, IEEE J. MEMS, 2000, 7, 1, 128-135.

[24] D. A. Skoog, J. J. Leary, "Principles of Instrumental Analysis", 4th Edition ed., Saunders College Publishing, Orlando, 1992, 8.

[25] A. Heilig, N. Barsan, U. Weimar, M. Schweizer-Berberich, J. W. Gardner, W. Göpel, Gas Identification by modulating temperatures of SnO2 based thick film sensors, Sens. Actuators B, 1997, 43, 45-51.

132 C H A P T E R 7

133

M. Heule, L. J. Gauckler,Sens. Actuators B, 2002, in press.

Miniaturised Tin Oxide Gas Sensors on Microhotplates by Micromolding in Capillaries

The concept of producing one gas sensor on one microhotplate is extended towards a gas sensor array consisting of several sensors on one hotplate. Twelve miniaturised gas sensors of nanoparticulate tin oxide were inte-grated as an array on a single microhotplate by using soft lithography, i.e. micromolding in capillaries. Soft lithography is a microfabrication tech-nique based on elastomeric molds of poly-dimethylsiloxane (PDMS) for pattern transfer. It is ideally suited to directly micro-shape liquid materi-als, including suspensions of oxidic powders. PDMS molds containing recessed microchannels could be placed directly - and removed after pro-cessing - on freestanding silicon nitride membranes of 250 nm thickness without causing damage. Tin oxide microlines were generated after apply-ing a suspension droplet at the entrance of the microchannels and filling them by capillary force action. A single sensor was shrunk on a minimal area of 10 by 30 µm2. In gas sensitivity experiments with carbon monox-ide concentrations from 100 to 2000 ppm, a maximal sensitivity to car-bon monoxide of 600 ppm was obtained. The heating power required to operate the whole array is reduced by at least one order of magnituede compared to the power that is usually required to run an array of multiple (> 10) micromachined gas sensors.

8.

134 C H A P T E R 8

8.1. Introduction

The ideal sensor is small, reliable, stable, low priced and operates on a mini-mum amount of energy. In semiconductor gas sensing applications based ontin oxide or other metal oxides, the introduction of microhotplates hasenabled researchers to come reasonably close to that ideal situation already.[1]

Sensor arrays are often employed to improve selectivity problems associatedwith metal oxide gas sensing or to probe more complex gas mixtures in elec-tronic nose applications.[2,3] Using micromachined hotplates, arrays can con-veniently be prepared even with a large number of sensors if needed.[4] Themost outstanding feature of microhotplates is the low amount of energy inthe range of 5 - 100 mW required to heat the active layer up to its sensingtemperature.[5] Based on microfabrication technology, microhotplates usuallyconsist of a heater and electrodes integrated on a freestanding thin-film mem-brane which has a low thermal mass of its own and is thermally isolated fromthe surrounding frame. The gas sensing layer is either deposited by thin-filmtechnology, such as sputtering,[6] chemical vapor deposition[7] or laser ablat-ing,[8] or by thick-film coating based on sol-gel processing or dispersions ofpowders.[9] It is also well established that thick-film coatings of nanoscaledpowders offer better sensing properties than thin films.[10,11] Using powdersynthesis routes, a better control over grain size, dopant stoichiometry andother parameters influencing the gas sensing performance is possible.[12]

Therefore, thick-film processes like droplet coating[13] and screen print-ing[14,15] were also used to coat microhotplates with a tin oxide layer. How-ever, these layers typically extend over areas of approximately 0.5 x 0.5 mm2.To our knowledge, the smallest tin oxide droplets deposited on microhot-plates have a dimension of 75 x 75 µm2.[16]

In this article, we present the use of soft lithography to fabricate very smalltin oxide lines of only 10 µm width directly on a microhotplate substrate.Together with platinum electrodes these lines form an array of 12 tin oxidegas sensors on a single microhotplate. Using this approach, the integrationdensity of tin oxide sensors reaches new dimensions of 200 gas sensing ele-ment per mm2 (calculation based on an area of 50 by 100 µm2 per sensor).Additionally, integrating more than one sensor on a single hotplate allows tocut energy consumption accordingly, e.g. 1/12 compared to operating a 12-

G A S S E N S O R S O N M I C R O H O T P L A T E S 135

hotplate sensor array. Soft lithography was developed as a simple way to pat-tern liquid materials in microfabrication.[17,18] Micropatterns are transferredby casting a silicone rubber, poly-dimethylsiloxane (PDMS), against a posi-tive relief structure (master) containing photoresist lines. The PDMS is thenpeeled off, cut and used as a mold that forms microcapillaries on a substratewhich can be filled with a liquid. This technique is referred to as Micromold-ing in Capillaries (MIMIC). The master structures may be reused manytimes to cast PDMS molds. Performing MIMIC does not require clean roomconditions once master structures have been fabricated. Dispersions of nano-particulate tin oxide in water were used, thus retaining the advantages ofpowder synthesis and thick-film technology in spite of the small size almostreaching thin-film dimensions. For the present study, the tin oxide nanopow-ders were undoped. However, the process can be transferred to other, morespecialised oxide powders.

8.2. Experimental

8.2.1. Microhotplate fabrication

100 mm silicon wafers of 450 µm thickness, polished on both sides (Wafer-net Inc., San Jose, CA, USA) were coated with a thin film of 250 nm low-stress silicon nitride in a LPCVD process (800°C, SiH2Cl2, NH3, by PSIWürenlingen, Switzerland). The electrode and heating coil pattern wasdefined photolithographically using Shipley 1813 positive resist (ShipleyInc., Marlborough, MA, USA) with a toluene soaking step before develop-ment to generate undercut resist edge profiles. Metals were deposited by e-beam evaporation of 5 nm Cr followed by 120 nm Pt. Subsequently, lift-offwas done using Microposit 1165 remover (Shipley Inc.) solution at 60°C. Forall dry etching processes, the Pt layer was protected by spinning another layerof Shipley 1813. The wafer was subsequently glued onto a supporting waferof the same size, again using Shipley 1813 photoresist as adhesive. On thebackside, the membrane opening pattern including dicing channels wasaligned and exposed in AZ4562 thick-film resist (AZ-Series, ClariantGmbH, Wiesbaden, Germany) using a Suss MA-6 mask aligner. First, thenitride and native silicon oxide were removed by reactive ion etching RIE.Secondly, deep etching DRIE through the whole wafer thickness was per-

136 C H A P T E R 8

formed in an STS Multiplex ICP system (STS Surface Technology Systems,Newport, UK). The wafers were separated by soaking the wafer stack in ace-tone over night and the single dies were obtained by breaking the wafersalong the precut dicing channels.

Fig. 8–1. Microhotplate fabrication scheme. a) coat 250 nm Si3N4 on silicon wafer by LPCVD. b) lift-off for structuring Pt electrodes. c) align backside pattern: membrane openings and dicing grooves. d) RIE dry etch to remove Si3N4 and ICP deep etching to release mem-brane. e) break wafer into dies, perform MIMIC on membrane. f) contacting by wire bonding, annealing on-chip.

silicon

photoresistSi3N4

a d

Pt PtPDMSSnO2

b e

c f) SnO2 microlines

G A S S E N S O R S O N M I C R O H O T P L A T E S 137

8.2.2. Micromolding in capillaries

The master containing a positive relief of the mold structures was preparedby standard photolithography. 9 g of Sylgard 184 PDMS prepolymer/catalystmixture (Dow Corning Inc., Midland, MI, USA) were poured over a masterstructure and cured for 26 hrs. The PDMS was peeled off and cut to pieces,in a way that the line pattern was cut open and became accessible for liquidson both opposite edges. Then, the PDMS molds were oxygen plasma-treatedfor 2 min using a 100 W Harrich PDC-32G sterilizer (Harrich Scientific,Ossinning, NY, USA). A plasma-treated PDMS mold was aligned perpendic-ular to the electrode pattern of a microhotplate chip under a stereo micro-scope and brought into contact. 10 µl of tin oxide suspension (nano-scaledSnO2 particles of 0.015 µm diameter, 15 wt% in H2O, as obtained from AlfaAesar) were dispensed at the entrance of the structures. After the capillary fill-ing, the sample was left to dry at room temperature. After 30 min, thePDMS mold was gently lifted with tweezers. Finally, the microsensor wasmounted on a printed circuit board (PCB) and contacted by wedge-wire-bonding using on a Delvotec 5425 (F&K Delvotec GmbH, Ottobrunn, Ger-many) semi-automatic wire bonding system. The PCBs were fitted into arubber-sealed 11 ml chamber through which the test gas mixtures wereinjected.

8.2.3. Gas sensitivity measurements

Digital mass flow controllers (El-Flow Digital, Bronkhorst Hitec, Ruurlo,Netherlands) were used for the mixing ratios and to control the total gas flow.Filtered air containing approximately 40% relative humidity was used as areference atmosphere. The analyte gases were purchased as certified mixturesof 2000 ppm in high purity nitrogen (PanGas AG, Dagmersellen, Switzer-land). Depending on the desired dilution ratio, the total gas flow rate was setbetween 150 ml min-1 and 300 ml min-1. In most cases the flow rates werefixed at 200 ml min-1, allowing to exchange the gas volume of the chamberwithin a few seconds. All measuring electrode leads were collected in a multi-channel thermocouple switch system 7001/7014 from Keithley. The electri-cal resistances of the tin oxide micro-sensors were recorded with a Keithley617 electrometer, the heating power and the resistive temperature device

138 C H A P T E R 8

(RTD)-signal were monitored using conventional digital multimeters, allconnected via GPIB to a PC. The heater power was delivered by standard4.5V radio batteries. The batteries were connected to the heater via a simpleself-made electronic circuit featuring a condensator (1 µF) in parallel to theheater and a potentiometer for voltage control. In a typical experiment, foursensors were monitored simultaneously by switching the electrodes every 2 to5 s.

SEM images were taken on a LEO 1530 (LEO Electron Microscopy Ltd.,Cambridge, UK) scanning electron microscope at low acceleration voltages(1 - 4 kV).

8.3. Results and Discussion

A new microhotplate design had to be implemented for this project, since nopreviously available designs featured multiple electrode pairs per plate. Thestandard interdigitated electrode geometries were not suitable either - theywould only have allowed for one miniaturized gas sensor. The layoutdesigned is shown in Fig. 8–2. For simpler microfabrication processing, a sin-gle layer design was chosen (so-called horizontal design), i.e. the heating coilsand the measurement electrodes were on the same mask. Therefore, only onemetallization/lift-off process had to be performed instead of two plus thedeposition of an insulating layer. There are two meandering heating coils(top to bottom), separated in the center by a simple line serving as a resis-tance temperature detector (RTD). Platinum has a very well defined positivetemperature coefficient of 3.9 x 10-3 Ω (ΩK)-1, which is proportional to tem-perature over a wide range.[19] From each side, interdigitated with the heat-ing coils, six 10 µm wide measurement electrode pairs that are disconnectedin the center come in parallel spaced apart by 10 µm. Therefore, a maximumof 12 individually addressable tin oxide sensors could be realised. There arealso two additional wire loops of 10 µm width, which allowed to check themembrane condition. A backside opening size of 1.5 by 1.5 mm2 and the sil-icon nitride layer thickness was chosen according to designs published earlierby Briand et al.[13] Due to variations in the DRIE process and slower etchingrates in corners, the actual membrane sizes ranged between 1.2 and 1.4 mm

G A S S E N S O R S O N M I C R O H O T P L A T E S 139

Fig. 8–2. Microhotplate layout scaled from original CAD files to define pho-tomasks. a) overview over Pt layout featuring a total of 20 connec-tions including heater and RTD temperature sensor. The device size was set to 8 x 8 mm2 in order to get enough area to perform MIMIC by hand. The backside opening for the isolated membrane is 1.5 x 1.5 mm2. b) close-up of the membrane layout with heating coils and RTD connected from the top, while measuring elec-trodes come in from two sides.

diameter. The metallization thickness of 120 nm was deliberately chosen inorder to limit the surface roughness for the subsequent MIMIC deposition oftin oxide nano-powder suspensions.

In Fig. 8–3, a schematic diagram including two microscope images of theMIMIC coating process is given. The microchannels on the PDMS consistedof two bundles of five 10 µm lines spaced by 30 µm. The channel height of6-7 µm was defined by the master structure thickness. Masters were also pho-tolithographically fabricated, spinning AZ4562 at 3500 rpm. PDMS moldswere aligned to the microhotplate in a way that the two bundles of lines weresituated in parallel on either side of the central RTD strip. This geometryallowed for a maximum of four micro-tin oxide elements spanning the same

Heating coils

6 electrode pairs RTD sensor

alignment marks

Service electrodes

Freestanding Membrane Area1.5 x 1.5 mm2

a b

Total Size: 8 x 8 mm2

140 C H A P T E R 8

Fig. 8–3. Schematic of the micromolding in capillaries process. a) PDMS mold is aligned and placed directly on the microhotplate chip. PDMS establishes a conformal contact and the microchannels are sealed. Micrograph of PDMS channels in contact. b) a suspension droplet at the side outside of the membrane is applied, capillary filling begins. c) wait for the suspension to dry, microlines of tin oxide are depos-ited inside microchannels. d) remove PDMS mold gently to release the structures. The arrow marks the former edge of the PDMS mold, where a strip of excess tin oxide nanopowder remains.

G A S S E N S O R S O N M I C R O H O T P L A T E S 141

electrode pair. We refer to such an assembly of one electrode pair and theportions of the tin oxide microlines which span these electrodes, as a singlegas sensor of the 12-sensor array.

These are therefore connected in parallel mode which has the advantagesof built-in redundancy. A single element may crack during later annealing orsome microchannels might not fill completely during MIMIC. Additionally,it could prove interesting to compare gas sensing properties of a differentnumber of lines connected in parallel mode. The microhotplate substrate wasused from microfabrication without prior treatment and MIMIC was donedirectly onto the freestanding silicon nitride membrane. The most outstand-ing feature of PDMS as mold material is that it establishes a conformal con-tact with many surfaces.[20,21] This contact is reversible similar to the well-known Post-It notes. During contact, microlines in the PDMS mold and thesubstrate surface form well sealed microchannels. The elasticity of PDMS canaccomodate roughness up to around 100-200 nm. And most importantly,PDMS can be placed on extremely fragile substrates. PDMS molds could beplaced onto and removed from microhotplates several times without everbreaking the 250 nm silicon nitride membrane. If the inside surface of thePDMS is made hydrophilic by oxygen plasma treatment, capillary forces forwater-based liquids are created. For a simple model of the thermodynamics ofcapillary filling, a rectangular microchannel with square cross-section andthree walls of PDMS and one wall of substrate material is assumed. ∆G com-bined with Young’s equation can be expressed as[22]

(1)

where x is the capillary width/height, z the filling length, with indices desig-nating interaction between solid surfaces S (PDMS) and S1 (Substrate) andliquid phase L, or vapor phase V. For liquids exhibiting contact angles ofθ > 90°, the free energy will become positive and the capillaries will not befilled. If only the substrate is not readily wettable such as Pt for water, a netnegative ∆G still results. The water contact angle on Pt was determined to beθPt = 89 - 92°. Since cos(θPt) = 0, no contribution to capillary filling results.The PDMS wetting term in equation (1) dominates by a factor of 3. There-

∆G 3xz γSV γSL–( ) xz γS1V γS1L–( )+[ ]– xzγLV 3 θ θ1cos+cos( )–= =

142 C H A P T E R 8

fore, the channels fill spontaneously with aqeous liquid and obstacles such asthe Pt electrodes can be overcome. After filling the channels with tin oxidesuspension, the drying process causes the newly formed ceramic powder net-work to sediment onto the substrate. A more comprehensive study on thefilling and drying behaviour of various tin oxide suspensions in MIMIC hasbeen published elsewhere. A series of micrographs taken from the device afterthe carbon monoxide sensitivity experiments at different magnification scalesis shown in Fig. 8–4. Tin oxide lines of only 10-11 µm width were obtainedupon removal of the PDMS mold. As observed from Fig. 8–4a, thereremained one microchannel unfilled, another two stopped filling at a Pt elec-trode edge. Unlike other thin-film coatings prepared by sol-gel spinning,these lines exhibit typical maximum thicknesses of approx. 3 µm. The cross-sections are predominantly trapezoidal-shaped (Fig. 8–4c). Many cracks areobserved due to drying or annealing.

Fig. 8–4. a) overview micrograph of the device that was used for gas sen-sing experiments. The single electrodes are numbered according to their connection with the electrical switch system. b) SEM close-up of a single electrode with two SnO2 lines spanning them. This entity is a single sensor of the array. c) SEM close-up of the left-hand microline, revealing its microstructure.

G A S S E N S O R S O N M I C R O H O T P L A T E S 143

As expected from the line geometry, cracks predominantly occur perpendicu-lar to the long axis of the structure. Additional stresses are induced by thebending of the material over the electrodes which results in occasionalnanometer-sized cracks at the base of the structures. The nanopowder sus-pension of 15 wt% used for the present work has too low a solids loading tocompletely circumvent the problem of cracking. However, the typical lengthof crack-free regions are in the same range of 15-30 µm as the gap such a tinoxide line has to span between two electrodes. Therefore, the chances arehigh that out of four tin oxide elements spanning the same electrode, at leastone or two electrical connections remain unsevered.

8.3.1. Heater calibration

Thermal calibration of the heater was carried out by sprinkling small grainsof pure organic substances with well-defined melting temperatures (p-anisicacid 183°C, caffeine 235°C, anthrachinone 283°C) and visually observingthe power required for melting them. The resistance signal of the RTD plati-num strip was also recorded as a function of heating power, which is given inFig. 8–5. It shows the RTD resistance difference at a certain heating powerlevel to its resistance at ambient temperature, typically 300-305 Ω. Assumingthat the RTD resistance is strictly proportional to the mean temperature onthe microhotplate, the resulting temperature increases slower with increasingelectrical power due to heat losses. A RTD resistance coefficient of 9.5°C Ω-1

was obtained from evaluating the heating powers necessary for reaching themelting temperatures of the organic standards. 300°C were reached at a heat-ing power of approx. 50 mW.

144 C H A P T E R 8

Fig. 8–5. Signal of the membrane-integrated RTD temperature sensor as a function of applied electrical heating power. The difference signal to its state at ambient temperature is shown.

8.3.2. Gas sensitivity, electrical characterisation

Annealing was carried out on-chip during first gas sensitivity measurementsat typically 290-300°C. The device accumulated several hrs (24-48 h) ofannealing at that temperature and for short periods of time only, peaks at400°C. Initially, the resistances between the electrodes scattered from 108 to1010 Ω. The base resistances in air of active sensors converged to approxi-mately 1 GΩ after the annealing procedure described. When dealing withsuch small dimensions, higher resistances were to be expected. With ‘active’sensors, we refer to those electrodes that were responding to carbon monox-ide concentration variations in the dilution series runs. On another device,carbon monoxide sensitivity was recorded even at 10 GΩ. Fig. 8–6 shows theresistance signals from three single sensors (in Fig. 8–4 designated as no. 13,16 and 23) as recorded in dilution series ranging from 100 to 2000 ppm car-bon monoxide in air. The nominal temperature was varied between 160°Cand 290°C in four runs. A fourth sensor, no. 26, was inactive. For all foursensors, data was acquired at a fixed temperature in the same run by switch-ing between the sensors every 3.3 s.

70

60

50

40

30

20

10

0

∆R

TD /

Ω

120100806040200Heating power /mW

∆RTD hotplate 1 ∆RTD hotplate 2 Polynomial Fit 1 Polynomial Fit 2

G A S S E N S O R S O N M I C R O H O T P L A T E S 145

Fig. 8–6. Raw resistances as recorded in gas sensitivity experiments. See text for details.

Sensors 13 and 16 consisted of two tin oxide lines spanning the electrodes,whereas sensor 23 was spanned only by one. That resulted in a slightly higherresistance of no. 23 throughout the experiments. Carbon monoxide pulses inthe dilution series were injected for 4 min, then the gas composition wasturned back to air for 5 min. If the detection limit is considered to beroughly three times the noise level, carbon monoxide could be successfullydetected starting at concentrations as low as 600 ppm and higher (sensor sig-nals 16, 23 overlapping at 160°C). This is consistent with earlier work on gassensors fabricated from pure tin oxide microstructures based on a larger pow-der grain size (250 nm).[23] The highest sensitivity was observed at 160°C.The last carbon monoxide peak at 2000 ppm exceeds the others by an orderof magnitude due to the fact that the gas mixture was changed considerably.It consisted only of the 2000 ppm mixture as purchased, containing only car-bon monoxide and nitrogen. In the absence of oxygen therefore, the sensitiv-

23

1613

108

2

3

4

567

109

2R

esis

tan

ce /Ω

100

1000

2000

CO

flow

/pp

m

500040003000200010000time /s

100

1000

2000

CO

flow

/pp

m

108

2

3

4

567

109

2

Res

ista

nce

/Ω

500040003000200010000time /s

23

16

13

23

16

13

23

1613

T = 120˚C T = 160˚C

T = 220˚C T = 290˚C

a b

c d

146 C H A P T E R 8

ity seems to increase considerably. However, bearing in mind that standardapplications for resistive gas sensors are usually performed at ambient air con-ditions, this way of sensitivity ‘improvement’ was not further investigated.

8.4. Summary

Miniaturisation of tin oxide gas sensors does not need to be limited by thedesign and size of the micromachined substrate or by the resolution of theoxidic material deposition method such as droplet coating or ink-jet printing.Novel coating methods, MIMIC and soft lithography offer easy prototypingmethods to shrink the size of the gas sensing element even more. Arrays of 12nanostructured tin oxide gas sensors on a single microhotplate have been pre-pared, each covering an area of only 10 by 30 µm2. Concentrations of600 ppm carbon monoxide in air were successfully detected using undopednanoscaled tin oxide. We believe that there is plenty of room for performanceimprovement. Particularly, there is the possibility to create miniaturised gassensor arrays often termed electronic noses, if e.g. the single sensors are dopedindividually or consist of different semiconducting oxides. We also see muchpotential in basic research on the gas sensing mechanisms of tin and otheroxides, e.g. to study the interactions of sensors in extremely close proximity.At IMCS 8, a paper on that topic already was presented by Wheeler et al.[24]

They examined the crosstalking effects between microhotplate sensors spaced300-400 µm, a distance which could be reduced considerably. In application,devices based on such designs may become important where the energy topower a sensor array is critical. Or, in cases where the space available for thegas sensor is extremely limited, such sensors could be integrated as a compo-nent into other MEMS devices performing a more complex task than gassensing only, e.g. micro gas reformers, micromachined fuel cells, space appli-cations or even components of solid oxide fuel cells may be shaped by thistechnique.

G A S S E N S O R S O N M I C R O H O T P L A T E S 147

8.5. References

[1] J. S. Suehle, R. E. Cavicchi, M. Gaitan, S. Semancik, Tin Oxide Gas Sensor Fabri-cated Using CMOS Micro-Hotplates and In-Situ Processing, IEEE Electron Device Lett., 1993, 14, 3, 118-120.

[2] H. T. Nagle, R. Gutierrez, S. S. Schiffman, The HOW and WHY of electronic noses, IEEE Spectrum, 1998, 9, 22-34.

[3] L. Ratton, T. Kunt, T. McAvoy, T. Fuja, R. Cavicchi, S. Semancik, A comparative study of signal processing techniques for clustering microsensor data (a first step towards an artificial nose), Sens. Actuators B, 1997, 41, 105-120.

[4] S. Semancik, R. E. Cavicchi, M. C. Wheeler, J. E. Tiffany, G. E. Poirier, R. M. Walton, J. S. Suehle, B. Panchapakesan, D. L. DeVoe, Microhotplate platforms for chemical sensor research, Sens. Actuators B, 2001, 77, 1-2, 579-591.

[5] G. Faglia, E. Comini, A. Cristalli, G. Sberveglieri, L. Dori, Very low power con-sumption micromachined CO sensors, Sens. Actuators B, 1999, 55, 140-146.

[6] I. Gracia, J. Santander, C. Cane, M. C. Horrillo, I. Sayago, J. Gutierrez, Results on the reliability of silicon micromachined structures for semiconductor gas sensors, Sens. Actuators B, 2001, 77, 409-415.

[7] C. J. Taylor, S. Semancik, Use of microhotplate arrays as microdeposition sub-strates for materials exploration, Chemistry of Materials, 2002, 14, 4, 1671-1677.

[8] M. A. El Khakani, R. Dolbec, A. M. Serventi, M. C. Horrillo, M. Trudeau, R. G. Saint-Jacques, D. G. Rickerby, I. Sayago, Pulsed laser deposition of nanostructured tin oxide films for gas sensing applications, Sens. Actuators B, 2001, 77, 1-2, 383-388.

[9] R. E. Cavicchi, R. M. Walton, M. Aquino-Class, J. D. Allen, B. Panchapakesan, Spin-on nanoparticle tin oxide for microhotplate gas sensors, Sens. Actuators B, 2001, 77, 1-2, 145-154.

[10] N. Barsan, M. Schweizer-Berberich, W. Göpel, Fundamental and practical aspects in the design of nanoscaled SnO2 gas sensors: a status report, Fresenius J. Anal. Chem., 1999, 365, 287-304.

[11] G. Martinelli, M. C. Carotta, E. Traversa, G. Ghiotti, Thick-Film Gas Sensors Based on Nano-Sized Semiconducting Oxide Powders, MRS Bulletin, 1999, 24, 6, 30-36.

[12] A. Chiorino, G. Ghiotti, F. Prinetto, M. C. Carotta, D. Gnani, G. Martinelli, Preparation and characterization of SnO2 and MoOx-SnO2 nanosized powders for thick-film gas sensors, Sens. Actuators B, 1999, 58, 01. Mrz, 338-349.

[13] D. Briand, A. Krauss, B. van der Schoot, U. Weimar, N. Barsan, W. Gopel, N. F. de Rooij, Design and fabrication of high-temperature micro-hotplates for drop-coated gas sensors, Sens. Actuators B, 2000, 68, 1-3, 223-233.

[14] D. Vincenzi, M. A. Butturi, M. Stefancich, C. Malagu, V. Guidi, M. C. Carotta, G. Martinelli, V. Guarnieri, S. Brida, B. Margesin, F. Giacomozzi, M. Zen, A. Vasiliev, A. V. Pisliakov, Low-power thick-film gas sensor obtained by a combina-tion of screen printing and micromachining techniques, Thin Solid Films, 2001, 391, 288-292.

[15] D. Vincenzi, M. A. Butturi, V. Guidi, M. C. Carotta, G. Martinelli, V. Guarnieri, S. Brida, B. Margesin, F. Giacomozzi, M. Zen, G. U. Pignatel, A. A. Vasiliev, A. V. Pisliakov, Development of a low-power thick-film gas sensor deposited by screen-

148 C H A P T E R 8

printing technique onto a micromachined hotplate, Sens. Actuators B, 2001, 77, 1-2, 95-99.

[16] P. P. Tsai, I. C. Chen, C. J. Ho, Ultralow power carbon monoxide microsensor by micromachining techniques, Sens. Actuators B, 2001, 76, 1-3, 380-387.

[17] Y. Xia, G. M. Whitesides, Soft Lithography, Angew. Chem. Int. Ed., 1998, 37, 550-575.

[18] Y. Xia, J. A. Rogers, K. E. Paul, G. M. Whitesides, Unconventional Methods for Fabricating and Patterning Nanostructures, Chem. Rev., 1999, 99, 1823-1848.

[19] National Instruments, Measuring Temperature with RTDs - a Tutorial, NI Applica-tion Note, 1996, 046.

[20] B. Michel, A. Bernard, A. Bietsch, E. Delamarche, M. Geissler, D. Juncker, H. Kind, J. P. Renault, H. Rothuizen, H. Schmid, P. Schmidt-Winkel, R. Stutz, H. Wolf, Printing meets lithography: Soft approaches to high-resolution printing, IBM J. Res. Dev., 2001, 45, 5, 697-719.

[21] E. Delamarche, M. Geissler, H. Wolf, B. Michel, Positive microcontact printing, J. Am. Chem. Soc., 2002, 124, 15, 3834-3835.

[22] E. Kim., Y. Xia, G. M. Whitesides, Polymer microstructures formed by moulding in capillaries, Nature, 1995, 376, 581-584.

[23] M. Heule, L. J. Gauckler, Gas Sensors Fabricated from Ceramic Suspensions by Micromolding in Capillaries, Adv. Mater., 2001, 13, 23, 1790.

[24] M. C. Wheeler, J. E. Tiffany, R. M. Walton, R. E. Cavicchi, S. Semancik, Chemi-cal crosstalk between heated gas microsensor elements operating in close proximity, Sens. Actuators B, 2001, 77, 1-2, 167-176.

149

M. Heule, L. J. Gauckler,Mat. Res. Soc. Symp. Proc., 2002, 687, 69-74.

Increasing the Integration Density by Vertically Separating the Heater of the Microhotplate Design

In most MEMS applications, dust and particles are avoided with consid-erable endeavor. However, for many applications such as gas sensors, pow-ders of functional ceramics would often provide better performance than corresponding thin film layers. Specifically where the functionality is based on a chemical reaction on surfaces, nanoscaled powders with a high spe-cific surface area have proven useful. This is the case for tin oxide gas sen-sors. Soft lithography was used for the fabrication of powder-based ceramic microstructures. In this paper, we present the integration of small tin oxide microstructures with an effective gas-sensing area of 10 by 30 µm2 on a micro-hotplate substrate with buried heating structures. Processing steps to prepare a 20 sensor array on the microhotplates are presented and dis-cussed concerning processing sequence, sensitivity towards 1000 to 1500 ppm hydrogen and power consumption. Additionally, effects of grain growth due to on-chip annealing of the ceramic nanostructure were observed.

9.1. Introduction

Microfabricated semiconducting gas sensors consist of a sensitive ceramiclayer that is typically supported on a thin freestanding membrane that can beheated to temperatures up to 500°C.[1-3] The most significant advantages ofminiaturized designs besides their small dimensions are the low power con-sumption in the range of mW and quick response times. Among the materi-als often used are tin oxide (SnO2), tungsten oxide, titanium oxide and many

9.

150 C H A P T E R 9

others. The range of detectable analyte gases include reducible gases likehydrogen, carbon monoxide and ethanol. Therefore, target applications arewarning sensors in security-relevant situations as well as room climate moni-toring. As far as the gas sensing process is understood today, the topmostlayer of oxygen is removed by a catalytic combustion of the analyte gas on thesemiconductor surface. This catalytic processes enable the sensors to detectvery sensitively. However, selectivity towards a specific gaseous species in a gasmixture is often the limiting obstacle that prevents the application of theseresistive type gas sensors. To a certain extent, the selectivity may be tailoredby adding dopants to the ceramics. One possible solution is the use of anarray of several differently reactive sensors simultaneously. The combinationof their responses in multivariate analyses or pattern recognition techniquesallows for unambiguous detection of gases in mixtures.

In this paper we present the fabrication of an array of very small ceramicelements of tin oxide by Micromolding in Capillaries (MIMIC), a techniquederived from soft lithography.[4] Key element are stamps of silicone rubber(poly-dimethylsiloxane, PDMS) that contain recessed microline patterns. It ispossible to integrate up to 20 single gas sensors in the present design on onemicro-hotplate using our approach. Our microhotplate fabrication design isbased on previously published work by Briand et al.[5,6] Thus, the powerconsumption of the whole sensor array can potentially be reduced by anotherfactor of 20. The flexibility of the method is demonstrated by using self-syn-thesized tin oxide powders with grain sizes in the range of 8 - 25 nm. It wasshown that gas sensors with a layer of nano-scaled powders exhibit far bettersensitivities combined with reasonable long term stability.[7]

9.2. Experimental

Microhotplate fabrication (Fig. 9–1): a 100 mm silicon wafer double-sidecoated with a 250 nm layer of LPCVD silicon nitride was used in a four-mask photolithography process. A heating pattern with built-in resistancetemperature detector (RTD) structures was defined by deposition and lift-offof 120 nm platinum (e-beam evaporization). This layer was then covered by300 nm silicon oxide as insulating layer (e-beam evaporization in reactiveatmosphere). Contact pads to the heater layer were exposed by photolithog-

V E R T I C A L M I C R O H O T P L A T E D E S I G N 151

raphy and opened by etching the oxide layer in buffered HF (12.5%, etchtime 255 s). An electrode layer featuring 40 adressable electrodes was depos-ited using the same Pt lift-off process as for the heater layer. On the backsideof the wafers, squares of 1.5 mm side length to define the area of free-stand-ing membrane and dicing grooves along the edges of the device were defined.The backside alignment was done using a Suss MA-6 mask aligner. Patternedwafers were set up on a supporting wafer before etching first the nitride layerin an STS ICP etch process (STS Surface Technology Systems, Newport,UK), then cut through the whole wafer by bulk micromachining in an STSDeep RIE etching system to release the membrane. Fabrication yield: out of21 devices per wafer, 18 freestanding membranes could be freed and diced bysnapping the wafer along the precut grooves. The dies were mounted onprinted circuit boards and contacted by wedge-wire bonding.

SnO2 Nanopowders. For the synthesis of tin oxide nanopowders, a modi-fied procedure based on an acid-catalyzed emulsion precipitation method oftin chloride by Vacassy was employed.[8] Tin(IV) chloride was obtained fromFluka AG, Buchs, Switzerland. All solvents and reagents used were of analyti-cal grade and used as received. After the precipitation reaction took place, thewet hydrolysis product was cleaned by redispersing the product from the

Fig. 9–1. Microhotplate with buried heating structures.

152 C H A P T E R 9

organic reaction phase in 200 ml of water (Millipore 18 MΩ cm) and settlingin a centrifuge at 2000 rpm for 1 hr. After 5 repeated cleaning cycles, waterwas removed to yield a 15%wt suspension of nanoscale wet tin oxide whichwas used in MIMIC. Grain sizes from 8 to 10 nm were found.

Gas sensitivity measurements. The process of delivering heating power,switching and measuring single sensor resistances and gas mixing was auto-mated using a temperature controller, a 39-channel thermocouple switch sys-tem, a standard multimeter and digital mass flow controllers. All devices werecontrolled by a PC over the GPIB bus. The sensor was mounted in a 11 mlvolume gas flow chamber.

9.3. Results and Discussion

It could be shown that even when used on an extremely fragile substrate likethe freestanding silicon nitride membranes fabricated, PDMS stamps can beplaced and lifted several times without breaking the membranes. Moreover,the PDMS material readily seals the capillaries and even accomodates roughsurfaces in the submicrometer region like in our case the steps induced by thethin film platinum elements incorporated in the membrane. The microcer-amic lines were formed by applying a suspension droplet of 10µl at theentrance of the capillaries. The stamps were aligned under a stereomicroscopeusing a xyz-translation stage and placed in a way that the ceramic lines con-nect the platinum electrodes. Images of the nano-ceramic lines and the gassensors they form are given in Fig. 9–2. One single gas sensor covers an areaof only 10 by 30 µm2, increasing the integration density dramatically.

V E R T I C A L M I C R O H O T P L A T E D E S I G N 153

Fig. 9–2. Top: Microscope image of the lines deposited onto a microhot-plate. The freestanding membrane covers most of the picture area and has a thickness of only 800 nm (1.5 mm square). Bottom: Close-up SEM view of the tin oxide lines (vertical, 10 µm width) spanning two pairs of electrodes. The numbered boxes 1,2 denote addressable gas sensors

It was decided to use an easy method capable of assessing the temperatures onthe membrane surface directly. The melting of pure grains of organic crystalswith well defined melting temperatures (p-anisic acid 185°C, caffeine 235°Cand anthraquinone 285°C) was observed as a function of current or heatingpower respectively. Additional information was obtained by following the sig-nal of the in-membrane RTD platinum stripes. Its resistances varied linearlybetween 365 Ω and 410 Ω for 0 mW resp. 80 mW heating power.

154 C H A P T E R 9

Gas sensitivity experiments were run over a heating power range of 0 to45 mW, corresponding to temperatures up to approximately 290°C. A firsthydrogen pulse at a concentration of 1000 ppm was injected during 1 min,then switched to air, after 2 min another 1500 ppm H2 were added andfinally the sensor recovery was observed in air during 2 min. The resultingsensor responses are shown in Fig. 9–3. The measured resistance values werenormalized by the resistance in air. The relative signals between differentpairs of electrodes have not been changed. On the left, the signal of three dif-ferent channels (a-c) are shown. The resistance drops significantly uponhydrogen exposure, e.g. to 35% of its value in air for a concentration of1500 ppm (a). Channels b and c seem not to have adequate electrical contactto the tin oxide ceramic layer. A smaller drop in resistance (max. 20%) uponhydrogen exposure can also be found for the platinum electrodes alone. Insubsequent SEM analysis of the device, small cracks in the SiO2 layer andtherefore in the platinum electrodes were detected. Presumably, they wereinduced by the bending of the membrane during heating. An annealing pro-cess that stabilizes the powder network and involves grain growth wasobserved in the tin oxide layer. SEM analysis showed that grain size after3 hrs of experiments had increased to approx. 20 - 25 nm. This observationis in good agreement with previous calcination experiments. We are currentlyworking on an improved design to overcome the problem of fast device dete-rioration.

V E R T I C A L M I C R O H O T P L A T E D E S I G N 155

Fig. 9–3. Initial gas sensitivity data for hydrogen at 225°C. Rectangular curves are showing hydrogen gas flow. a, b, c are measured on dif-ferent electrodes (channels).

Miniaturizing ceramic gas sensors to the ten-micrometer range seems possi-ble. Such small, highly integrated resistive type gas sensors based on suspen-sions of nanoscaled powders are feasible by using novel microdepositionroutes such as Micromolding in Capillaries. The PDMS stamps (i.e. softlithography) are very useful to microshape not only colloidal suspensions butall types of liquid materials even on extremely fragile micromachined siliconnitride membranes of less than one micron thickness. Future work will focuson improving the electrical reliability and on making full use of array capabil-ities of such microhotplate-based gas sensors.

a

b

c

1.5

1.0

0.5

0.0

R /

R0

400300200100time /s

1000

1500

H2 flo

w /p

pm

156 C H A P T E R 9

9.4. References

[1] J. S. Suehle, R. E. Cavicchi, M. Gaitan, S. Semancik, Tin Oxide Gas Sensor Fabri-cated Using CMOS Micro-Hotplates and In-Situ Processing, IEEE Electron Device Letters, 1993, 14, 3, 118-120.

[2] S. Semancik, R. E. Cavicchi, M. C. Wheeler, J. E. Tiffany, G. E. Poirier, R. M. Walton, J. S. Suehle, B. Panchapakesan, D. L. DeVoe, Microhotplate platforms for chemical sensor research, Sens. Actuators B, 2001, 77, 1-2, 579-591.

[3] G. Faglia, E. Comini, A. Cristalli, G. Sberveglieri, L. Dori, Very low power con-sumption micromachined CO sensors, Sens. Actuators B, 1999, 55, 140-146.

[4] Y. Xia, G. M. Whitesides, Soft Lithography, Angew. Chem. Int. Ed., 1998, 37, 550-575.

[5] D. Briand, A. Krauss, B. van der Schoot, U. Weimar, N. Barsan, W. Gopel, N. F. de Rooij, Design and fabrication of high-temperature micro-hotplates for drop-coated gas sensors, Sens. Actuators B, 2000, 68, 1-3, 223-233.

[6] D. Briand, B. v. d. Schoot, N. F. d. Rooij, H. Sundgren, I. Lundström, A Low-Power Micromachined MOSFET Gas Sensor, IEEE J. MEMS, 2000, 9, 3, 303-308.

[7] M. C. Carotta, M. Ferroni, D. Gnani, V. Guidi, M. Merli, G. Martinelli, M. C. Casale, M. Notaro, Nanostructured pure and Nb-doped TiO2 as thick film gas sensors for environment gas monitoring, Sens. Actuators B, 1999, 58, 3, 310-317.

[8] R. Vacassy, R. Houriet, G. J. G. Plummer, J. Dutta, H. Hofmann, "Tin Dioxide Nanopowders for Gas Sensor Applications", edited by K. E. Gonsalves, M. I. Bara-ton, R. Singh, H. Hofmann, J. X. Chen, J. A. Akkara, Materials Research Society, MRS Res. Soc. Proc., 501, Warrendale PA, 1998, 41-46.

157

Validating the Concept of Miniaturising Resistive-type Gas Sensors

In the field of semiconducting gas sensors, researchers from the Institute of Physical and Theoretical Chemistry of the University of Tübingen (IPTC) have acquired many years of experience. The concept of miniaturizing metal oxide gas sensors using soft lithography could be validated in collab-oration with Dr. N. Barsan, Dr. U. Weimar and Dr. A. Gurlo of IPTC. Gas sensors were prepared with MIMIC-microlines of In2O3 powder on substrates from IPTC. Sensitivity measurements with CO, Ethanol, O3 and NO2 were performed in Tübingen.

10.1. Experimental

A quantity of 150 mg In2O3 powders (synthesis by Dr. A. Gurlo, IPTC, usedas received, ρ = 7.13 g/ml, IEP = 8-9) with particle sizes of approx. 25 nmwere dispersed in 300 µL acetate buffer solution at pH 4.5 which contained82 mg NaOAc (10 mM), 36 mg HOAc (10 mM) dissolved in 60 g H2O.The resulting dispersions of 6.5 vol% were sonicated with an ultrasonicprobe at 400 W (VibraCell VC-750, Sonics Inc., Newton, CT, USA).MIMIC experiments were performed by adding a 7-10 µL droplet at theentrance of PDMS capillaries on Si wafer substrates. For later dispersions,6 µL of polyacrylate dispersing agent were added to the composition. Prepa-ration: 1.84 g polyacrylic acid sodium salt 2100 dissolved in 6 g water, 0.06 gNH3 (25% in H2O) added, adjusting the pH to 9.9.

10.

158 C H A P T E R 1 0

MIMIC was performed on sapphire sensor substrates (‘Eurochips’) with thinfilm metallisations (Pt, 400 nm) forming interdigitated electrodes for resis-tance measurement on the front, heater and temperature sensing (RTD)structures on the backside. PDMS mold alignment was done manually whileobserving with a stereo microscope. The coated Eurochips were thenmounted in standard TO-12 packages by electrodischarge bonding of Ptwires (300 µm).

The sensors were wired and mounted in a Teflon chamber and connectedto a gas delivery system capable of supplying well defined gas mixtures of dif-ferent analytes in synthetic wet (pO2 = 0.2, pN2 = 0.8; 50% rel. humidity)and synthetic dry air. Several series of gas sensitivity measurements with pre-conditioning times of 120 min were run. Analyte gas was injected typicallyfor 30 min before switching back to wet/dry synthetic air. Sensitivities forCO, C2H5OH, O3 and NO2 were tested. The ppb levels of O3 and NO2 gasconcentrations were cross-checked after passing the sensors by independentgas analyzers.

Several own microhotplate sensor channels of SnO2 (Alfa Aesar) on‘mhp2’ substrates were also tested for CO and C2H5OH in a separate hous-ing connected the same gas flow.

10.2. Results and Discussion

MIMIC-coated sapphire Eurochips are shown in Fig. 10–1. In dispersionpreparation, it turned out that pH-buffering was sufficient for obtaining sta-ble suspensions. However, the green bodies adhered badly on the substrate toan extent that the microlines were usually destroyed when lifting the PDMSmold. The addition of poly-acrylic acid that was used as a dispersing agentfor tin oxide made the green lines more stable in the case of indium oxide.Care had to be taken that the remainders of the droplets applied for capillaryfilling did not short-circuit the electrodes. Fig. 10–1a shows a clean samplewhere only MIMIC lines span the interdigitated electrodes, whereas the resis-tance of a partially short-circuited sample was significantly lower (Fig. 10–1b). It was estimated that some the microlines span the electrodes at 50 dif-ferent locations with varying thicknesses and powder coverages. The geome-try of such a single MIMIC-line is 10 by 100 µm2.

V A L I D A T I O N E X P E R I M E N T S 159

Fig. 10–1. a) overview on the sapphire-Pt sensor chip coated with MIMIC lines. Due to its transparency, the backside layer containing heater and temperature sensing structure is also visible. b) SEM overview on a typical sample of MIMIC indium oxide microlines. The powder-covered area results from the droplet at the side of the PDMS mold. c) SEM close-up of an indium oxide microline pattern.

The results of gas sensitivity measurements with CO and C2H5OH repre-senting reducing gas analytes are shown in Fig. 10–2. Three MIMIC coatedand one screen-printed reference samples were measured simultaneously. Theresistances reflect the amount of coating present. Therefore, the referencesamples exhibited the lowest resistance (dotted line), then came the samplesthat were partially short-circuited by non-MIMIC structures, and finally, thesample where only MIMIC lines contributed to the resistance. Indium oxideshows good sensitivity towards C2H5OH and almost none to CO, at250 ppm levels. The drop in resistance upon gas exposure is higher in thecase of MIMIC-lines, by roughly a factor of 2. This effect is not sopronouced under dry air conditions, where the resistance drop is similar.However, the baseline is not that stable and drifts during the measurement.

160 C H A P T E R 1 0

Fig. 10–2. Responses of indium oxide microlines (thick line) compared to a reference screen-printed sensor (dotted line). Weak lines show sensors covered with MIMIC-lines but short-circuited to a certain extent by the remainders of the side droplet (see Fig. 10–1b).

In the case of oxidizing gases, useful signals could be recorded as well. Sinceoxidizing gases enhance the space-charge layer by depositing oxygen on theoxide surface, the resistances increase upon gas exposure very sensitively. Lev-els of ppb concentrations are sufficient to induce a good signal. The resultsshown in Fig. 10–3 exhibit a well resolved peak even for concentrations aslow as 30 ppb of O3. NO2 levels of 300 ppb could also be clearly resolved. Incontrast to the situation with reducing gases, it seems that the MIMIC-linesensitivity at higher resistance levels is lower than that of the reference sensor.Care in interpretation is advised, since the levels reach the 1010 Ω level atwhich resistance measurement becomes more less quantitative and the S/Nincreases.

103

104

105

106

Res

ista

nce

/Ω

10008006004002000Time /min

0

250

500

CO

, EtOH

/pp

m

CO Et. CO Et. CO Et. CO Et. CO Et. CO Et.

dry air

V A L I D A T I O N E X P E R I M E N T S 161

Fig. 10–3. Response of In2O3 sensors towards oxidizing gases O3 and NO2. a) signals from MIMIC lines and the reference sensor. b) signals from the layer- and MIMIC-coated versions.

10.2.1. MIMIC Sensors on Microhotplates

The gas mixing apparatus was furbished with analyte gas mixtures of500 ppm CO/C2H5OH. An air flow at 50% rel. humidity with a maximumanalyte concentration of 250 ppm was prepared by mixing equal parts ofhumidified synthetic air (100% r.h.) with the analyte gas mixture. As previ-

102

103

104

105

106

107

108

109

1010

Res

ista

nce

/ Ω

10008006004002000

0100300500

750

1000

O3 , N

O2 /p

pb

O3

30 ppbNO2

50 ppb

In2O3 microlines reference sensor

102

103

104

105

106

107

108

109

1010

Res

ista

nce

/Ω

10008006004002000Time /min

0100300500

750

1000

O3 , N

O2 /p

pb

O3

30 ppbNO2

50 ppb

a

b

162 C H A P T E R 1 0

ous experiments had shown, the microsensors prepared from undoped SnO2were not significantly sensitive to concentrations lower as 500 ppm of CO.The same behaviour was found in these measurements. In only one instance,a weak signal could be recorded during exposure to 250 ppm CO (Fig. 10–4). A resistance drop of 4.3% out of a 162 MΩ level was recorded.

Fig. 10–4. MIMIC-line sensor signal for 250 ppm CO at a temperature of 280°C (47 mW heating power).

10.3. Summary

It could be confirmed that the observed sensitivity limit of MIMIC lines arenot a consequence of miniaturization but of the oxide material employed.The materials versatility of the micromolding processes has been demon-strated by using indium oxide powders. Colloidal dispersions of alternatepowders could be prepared within a short time. A big advantage of the col-laboration and support from IPTC was the possibility for a direct compari-son of sensing properties between microline-coated versions with screen-printed versions of the same powders on same substrates. Shrinking the gassensing elements in size to approx. 50 single line portions spanning the elec-trodes (each 10 by 100 µm2) did not adversely affect gas sensing properties of

170

165

160

155

150

Res

ista

nce

/MΩ

806040200Time /min

0

250

CO

/pp

m

CO

V A L I D A T I O N E X P E R I M E N T S 163

the material. In some cases, even a slightly increased sensitivity was obtainedat correspondingly higher resistance levels. For oxidizing gases, concentra-tions of 30 ppb ozone could be detected.

164 C H A P T E R 1 0

165

Outlook

11.1. Microstructures from Colloidal Systems

Integrating ceramic powders with microfabricated devices is possible. Pow-ders are a novelty in today’s microfabrication technology. A combination ofcolloidal dispersions and standard lithography or alternate microfabricationschemes like soft lithography was used to fabricate miniaturised gas sensorsand other devices. Many competing processes do not offer a true micrometerresolution. Ceramic microstructures with 10 µm features and reproductionaccuracies of 1-2 µm could be generated routinely. The size of the gas sensingelements of semiconducting metal oxides could be miniaturized by twoorders of magnitude while retaining the properties of a classical thick-filmcoating. These may have several advantages over vacuum-deposited thinceramic films in applications where high specific surface, controlled porosityor precisely controlled composition and morphology are of importance.

The photoresist casting method (PRC) allows a maximum freedom ofdesign geometry. While the lateral dimension is defined by standard photoli-thography, the height is given by the resist thickness. Thus, ceramic micro-structures with roughly the same resolution as the lithographic process can begenerated. Main drawback is the need for processing every sample in a cleanroom environment before casting the colloidal dispersion.

Micromolding in Capillaries (MIMIC) was used as a very fast method forgenerating microscaled lines featuring lengths up to a few mm. Alignment ona 5-10 µm scale is easy and straightforward. A clean room is not required.The wetting of surfaces by PDMS seals the capillaries effectively and due to

11.

166 C H A P T E R 1 1

its elasticity also accomodates a certain surface roughness of approximately500 nm. MIMIC was chosen as the main working horse during this thesiswork. Disadvantages are the difficulty to control the exact line length and apossible spoiling of the side areas by remains of the suspension. The methodin its current state is better suited for rapid prototyping in a laboratory thanfor processing larger amounts of samples.

In Microcontact Printing, the lateral structures are defined by the PDMSmold as well. The thickness of the powder coatings will vary laterally and is afunction of structure size, wetting behaviour of the dispersion and dryingshrinkage. However, the principle of µCP is the most simple of all theschemes considered in this work. This circumstance makes it particularyattractive to continue the development towards possible low-cost mass pro-duction schemes.

11.2. Prospects in Gas Sensing Applications

As has been demonstrated, several miniaturized gas sensing elements can beintegrated on one microhotplate. Up to now, the concept has been to buildsensor arrays by combining several hotplates on one chip which is easily pos-sible by slight adaptations of the mask designs. However, the power necessaryfor driving such a conventional microhotplate sensor array increases with thenumber of microhotplates. Therefore, the possibility to place the entire sen-sor array on only one microhotplate could be crucial in applications wherethe power available for the array is severely limited, e.g. space applications orsensors integrated in a wrist-watch setup.

All the opportunities offered by electronic nose applications can also beapplied to such miniaturised versions. It would be very interesting to deviseways to integrate two or more different sorts of powders on the same areawhich then could be doped with varying metals.

However, the versions of microsensors prepared for this thesis exhibitedlarge variations in electrical resistances and hence varying sensitivity proper-ties were found. In the case of microhotplate sensors, the exact nature of con-tact resistances could not be elucidated.

O U T L O O K 167

The sensitivity limits of 600 ppm CO and 100 ppm H2 obtained with thesegas sensors are inferior to those of existing gas sensors which can detect con-centrations of reducing gases down to the 1-100 ppm level. This is not a fun-damental problem of the miniaturisation approach. These sensitivity limitsare inherent to the undoped tin oxide used for micropatterning as could beshown during the validation experiments in Tübingen (reference sensitivitymeasurements in Appendix A4.3). As also has been extensively shown withdifferent powders varying from 10 to 250 nm particle size, all processes maybe adapted to more suitably doped tin oxide powders. It would have beenbeyond the scope of this work to enter the field of optimising powder synthe-sis routes for gas sensing.

On the scientific side, it remains unclear as to how the gas sensitivityscales with size. It is well established that sensitivity depends on the specificsurface area of the semiconductor. However, that does not put a lower limiton the size of the sensing layer. By taking nano-scaled powders, the sensitivitycould be increased. Further reduction in size of the sensing layer will requiresmaller electrode geometries. Additionally, the contact resistances of themetal electrodes to the oxide material need to be better characterized. Theseproblems can certainly be overcome. The double line approach presented inchapter 3 or smaller structures made by PRC might be used as a firstapproach for a further reduction of the size of the sensing layer. First experi-ments towards this end have already been published.[1] Semancik’s group atNIST measured the hydrogen sensitivity of sputtered tin oxide films as thinas 1.5 nm. Ultimately, one could ask the question whether a single nanopar-ticle could be used as a gas sensor as well. Other researchers have been think-ing about the same question as well.[2]

The sensing mechanisms of the catalytic sensing reactions are still notunderstood completely. One main problem is that analytical methods formeasurements under real sensing conditions on a particulate layer with anirregular microstructure are barely available. Thus, sensing layers with a cer-tain order, e.g. a single layer of particles or a single row of particles wouldprovide interesting reference sensors. Small sensing layers could be the key to

168 C H A P T E R 1 1

a well-known microstructure of limited complexity that simultaneously couldalso be modeled in a computer. These could be created on the basis of themicrofabrication methods presented in this thesis.

As confirmed with the evolving versions of miniaturized tin oxide gas sen-sors, the electrical resistance increases due to geometrical reasons. There arealso potentiometric type gas sensors based on solid state ionics. There, thesensor signal is a Nernstian voltage that should almost be independent on sizeeffects. Contacts to researchers in this field have been established.[3]

11.3. Ceramic Powders as Elements in MEMS technology

First steps towards integration with MEMS technology were demonstrated.Powder-based ceramic microstructures can be made clean-room compatible(Chapter 7). According to most engineers, alternate microfabricationschemes can only be performed before (pre-processing) or after (post-process-ing) clean room fabrication. Within this project however, sintered tin oxidemicrolines could be cleaned in an ultrasonic bath, taken back into the cleanroom and were used again for photolithography for the subsequent lift-offprocesses. The photoresist also acted as a protection layer against the loss ofparticles. The case of a typical post-processing application has been exampli-fied in Chapters 8 and 9 by depositing MIMIC lines onto extremely fragilethin film membranes of only 250 nm thickness.

In order to make the processes used in this thesis fully compatible withMEMS production, two large steps still have to be taken. First, the processwill have to be adapted for wafer-level processing, i.e. the parallel coating of anumber of devices on the same wafer must become practicable. The secondstep consists in applying the process to many wafers in succession. For thesescaling-up requirements, only µCP and PRC have the immediate potential.Large-area PDMS stamps can be made and, for example, applied onto arotating cylinder for consecutive printing. The next production step wouldinvolve dip-coating, as discussed in Chapter 4. The first experiments towardsthat direction have been presented by IBM Zurich’s research laboratories.[4]

Yet, the first implementation in a complete MEMS fabrication process stillneeds to be realised.

O U T L O O K 169

11.4. Combining Bottom-up and Top-down Approaches

At a first glance, methods from soft lithography seem to consist of purelytop-down microfabrication approaches. A closer look, however, reveals cer-tain aspects of bottom-up approaches as well, e.g. the spontaneous PDMSwetting of the surface or the ink transfer in µCP whose driving force is theSAM formation of ink molecules. Colloidal dispersions with particles thathave a self-assembling ability have been prepared before.[5] It could provevery interesting to combine a self-assembly process with liquid microtempla-tion so to generate rather complex micropatterned structures in just onedepositing step.

170 C H A P T E R 1 1

11.5. References

[1] G. E. Poirier, R. Cavicchi, S. Semancik, Ultrathin heteroepitaxial SnO2 films for use in gas sensors, J. Vac. Sci. Technol. A, 1993, 11, 1392-1395.

[2] S. Semancik (Personal Communication), International Meeting on Chemical Sen-sors IMCS 9, Boston, MA, USA., July 2002.

[3] H. L. Tuller (MIT Boston), E. Traversa (University of Rome) (Personal Communi-cation), 102nd Meeting of the Electrochemical Society, Salt Lake City, UT, USA., October 2002.

[4] B. Michel, A. Bernard, A. Bietsch, E. Delamarche, M. Geissler, D. Juncker, H. Kind, J. P. Renault, H. Rothuizen, H. Schmid, P. Schmidt-Winkel, R. Stutz, H. Wolf, Printing meets lithography: Soft approaches to high-resolution printing, IBM J. Res. Dev., 2001, 45, 5, 697-719.

[5] D. M. Dabbs, I. A. Aksay, Self-assembled ceramics produced by complex-fluid templation, Annu. Rev. Phys. Chem., 2000, 51, 601.

171

ppendices

A.1. Microsystem Design Notes

In this section, the design approaches and evolution of the microfabricationprocesses are discussed. All mask graphics are directly taken from theAutoCAD files that were the basis for photolithograpy mask fabrication inorder to retain the relative dimensions. Photomasks usually consist of a 5-inch glass plate coated with a UV-opaque chromium film which is patternedby direct laser-writing according to the CAD file. The services used for maskfabrication in this work were IMT Masken und Teilungen AG, Greifensee,Switzerland and DeltaMask, Enschede, Netherlands.

A

172 A P P E N D I C E S

A.1.1. Micro-Gas Sensors on Sapphire

The first version of micro-gas sensors was produced using a mask designedvery early into the thesis work. It contained various line patterns and twoelectrode pattern that matched each other in size. The sapphire plates had to

Fig. A–1. Mask elements to define all pattern on sapphire single crystals. a) heater pattern for Pt-lift-off. b) after the SiO2 deposition, the same pattern was partially covered with a piece of silicon wafer shaped as indicated in the schematic. Through the resulting openings, the access windows to the electrodes were etched using HF/NH4F. c) line pattern used for casting PDMS channels. d) electrode pattern matching the 10-line pattern c), which was defined in the end for the last Pt-lift-off process.

a b

c d

A P P E N D I C E S 173

be exposed singly and one by one for each photolithography step. The align-ment was done as follows: deposition of heater pattern centered on the sap-phire wafer, openings in the silicon oxide layer by adjusting to the edges ofthe squares (Fig. A–1a), MIMIC coating by eye perpendicular to the heaterpattern direction, finally, careful alignment by looking through the electrodepattern at the tin oxide microlines. Using the functional structures them-selves for aligning may complicate the practical work significantly. Undermicroscope view not all patterns are easy to recognise. Therefore, so-calledalignment marks were introduced in later designs.

A.1.2. Vertical Microhotplate Design

Microfabrication processes quickly become time-consuming and costly ifonly one device is done at a time. The power of microfabrication processeslies in the parallel processing of a multitude of samples. For fabricatingmicrohotplates, the substrate material was then switched to 100 mm siliconwafers. This allowed to design for 21 identical patterns on one mask or wafer.One exposure, deposition, etching, lift-off step etc. could be simultaneouslyapplied to all 21 devices. The size of a single hotplate was 17 by 17 mm2.

174 A P P E N D I C E S

Fig. A–2. First of four masks used in the vertical design of microhotplates: heater pattern.

The heater pattern consists of five contacts, two of them heating meanders,the other three straight and thin bars over the membrane area for RTD (resis-tive temperature device) temperature measurement. The Pt squares for wire-bond contacting were chosen to be generously large, 0.8 mm (one well-placed bond typically needs 50 µm). As shown in Fig. A–2c), the electricalleads remain broad (200 µm) until the freestanding membrane begins, thenthey are linearly reduced to 10% width (20 µm), which was also the width ofthe meandering heater pattern and the RTD stripes. Using such a geometry,the regions of highest electrical resistance in the film were placed directlyonto the membrane.

a) heater layer 100 mm mask b) single die

c) membrane close-up

A P P E N D I C E S 175

In the center of the future edge of each device and on both sides of the bond-ing pads for heating/temperature sensing, alignment crosses were placed. Theprinciple of alignment marks is shown in Fig. A–3. A cross-shape is of advan-tage, since its two symmetry axes go parallel the movement of the xy-axis ofthe mask-aligner. A cross is shaped on the substrate on the first mask, in thisexample integrated with the heater pattern. On the next mask, correspondingslightly enlarged target marks are included. If one is to coincide them, a smallgap d of typically 5 µm remains. Now, the human eye can easily be deceivedby parallel lines, but we have a fine sense for symmetry. So, it is quite easy tojudge whether the alignment of the cross within the square is symmetric. Anyshift x from the balanced center position is perceived or enhanced by a factorof 2: on the closer side, the gap becomes smaller d-x, whereas on the fartherside the gap enlarges by the same distance d+x. If done using two such marksacross a few cm on the wafer, an alignment accuracy of approximately 10-20% of the gap d may be reached which is about 1 µm for d = 5 µm.

Fig. A–3. Alignment marks, working principle.

mark on mask to expose mark on substrate/previous mask

well aligned slightly misaligned strongly misaligned

coincide

176 A P P E N D I C E S

Fig. A–4. Mask defining the openings in the silicon oxide layer for electri-cal contact to the buried heater layer.

a) open heater layer 100 mm mask b) single die

c) openings close-up

A P P E N D I C E S 177

Fig. A–5. Electrode pattern with numbered wire-bond pads.

In Fig. A–5, 4 electrode patterns seem to be missing. At these locations, pat-terns for PDMS casting and eventually, photoresist casting, were included.

a) electrodes 100 mm mask b) single die

c) membrane close-up

178 A P P E N D I C E S

Fig. A–6. Backside opening mask consisting of central squares 1.5 by 1.5 mm2. c) shows a view through the etched opening from the backside on the released freestanding membrane including all Pt layers.

In the final step, the silicon wafer was etched from the backside through itswhole thickness to release the stack of Si3N4, Pt, SiO2, Pt membranes. Themask for the backside openings contains a central square 1.5 by 1.5 mm2,alignment crosses and 60 µm wide lines along the device edges that willbecome grooves to cut out the single hotplate chip. There was no need for aspecialised wafer saw after backside etching, the devices were simply brokenalong those grooves. However, the wafer had to be glued to a supportingwafer prior to the etching process using a wax (Crystalbond 555, AremcoInc., Valley Cottage, NY, USA) that was melted at 60°C and could beremoved in acetone or hot water.

a) backside openings 100 mm mask b) single die

c) membrane close-up

A P P E N D I C E S 179

A.1.3. Horizontally Designed Microhotplates

Using only two masks, a horizontal design where all electrodes, heating pat-terns and RTD strips were placed on the same mask adjacent to each other,was implemented. Two advantages were important: First, that design allowedto fabricate approx. 80 microhotplates per wafer. Second, the fabrication pro-cess was reduced to one Pt-lift-off and the backside etching process. Althoughonly 12 electrode pairs could be integrated, sacrificing some integration den-sity, the availability of microhotplates no longer posed the limiting factor inpreparing micro-thickfilm gas sensors. For the completion of this thesis, onlytwo wafers in total had to be processed with a yield of approx. 70%.

Fig. A–7. Frontside mask of horizontal microhotplate design integrating heaters, RTD strips and sensing electrodes on the same Pt-lift-off layer.

a) heater/electrode layer 100 mm mask b) single die

c) membrane close-up

180 A P P E N D I C E S

Fig. A–8. Backside openings including snapping grooves and alignment marks. c) would be the freestanding Pt/Si3N4 membrane as seen from the back through the opening.

For the backside pattern, the alignment marks were integrated seamlessly intothe snapping grooves (see Fig. A–8b, 2 at the top, 2 at the bottom). There-fore, only the two side arms remained.

a) backside opening mask 100 mm b) single die

c) membrane close-up

A P P E N D I C E S 181

A.2. Improved Clean Room Procedures

A.2.1. Photolithography

If repeated photolithographies are to be successful, it is crucial to maintain awhole set of parameters constant. All photolithography for the present workwas carried out in a class 10 clean room of IMES (Institute of MechanicalSystems, Prof. J. Dual, Dr. S. Blunier) at ETHZ at 40% relative humidityand a fixed temperature of 21-22°C. In parameter confirmation runs, thereproduction quality was constantly improved. For AZ4562 photoresist, theprocedure in Table A–1 was repeatedly and predominantly used.

Table A–1. Standard AZ4562 photolithography parameters.

ProcedurePositive resist AZ4562 (Clariant GmbH, Wiesbaden, Germany)

Cleaning Quick-dump rinse until 16 MΩ cm

Adhesive Coating hexa-methyl-disilazane HN(Si(CH3)3)2 vapour over night

Resist Spinning 2.7 ml for 4” Wafer, 500 rpm/3s to distribute, then accelerate at 1000 rpm s-1 to 3500 rpm, hold for 35 s.Leave to dry 5 min, resulting thickness 7 µm

Prebake 1. Oven 100°C, 30 min or2. Hotplate 100°C, 3 min

Exposure 700 mJ cm-2, typically 35 s at 20 mW cm-2, λUV = 250 nm

Development Microposit 351, 1 part diluted in 3 parts H2O (Shipley Inc., Marlborough, MA, USA)resulting in typical development times of 35-45 s, depending on how the samples were stirred during development

Bottom cleaning of resist pattern

Oxygen Plasma treatment(TePla 300E, Tepla AG, Feldkirchen, Germany)pRF = 220 W, pO2 = 85 mbar, t = 45 s

(Postbake) Oven, 120°C, 30 min

182 A P P E N D I C E S

A.2.2. Improving Pt Lift-off Quality

When processed under the conditions as stated above (see Chapter A.2),AZ4562 exhibited a strongly rounded sidewall profile. After Pt was evapo-rated onto those structures, the lift-off produced Pt structures with jaggededges. The Pt formed one contiguous film over the whole topography andtherefore ruptured unevenly during lift-off, see Fig. A–9. By soaking thewafer after exposure in toluene for 1 min, toluene intercalates a the first few100 nm of resist thickness and slows its dissolution rate during development.Once the developer passes the toluene-soaked region, dissolution rates speedup to standard pace. Now under the first layer, the dissolution also spreads alittle faster sideways, creating the desired undercut structure. These undercutsthen shadow the edge upon Pt evaporation leading to separated metal coat-ings on the resist layers, or on the wafer surface, respectively (see Fig. A–9b).There is no need for a rupture in the film during lift-off as in the situationfrom Fig. A–9a.

The better approach was to produce undercut photoresist structures, in away that the Pt film on top of the resist did not contact the Pt on the bottomof exposed resist structures. A simple process from Chalmers University Swe-den was found on the interneti, see Fig. A–2.

i. http://fy.chalmers.se/assp/snl/public/wproc/S1813Liftoff.html

A P P E N D I C E S 183

Table A–2. Undercut photoresist profile for Pt lift-off.

ProcedureShipley 1813 positive resist (Shipley Inc., Marlborough, MA, USA)

Spin coating 2.0 ml at 500 rpm/3 s, then 1000 rpm s-1 to 4000 rpm/35 sresulting thickness 1.6 µm

Prebake hotplate, 1 min, 100°C

Exposure 5 s at pUV = 16.2 mW cm-2 (dose 81 mJ cm-2)

Toluene soaking immerse 1 min in toluene, blow dry with N2 gun

Postbake hotplate, 15 s, 90°C

Development 16 s in Microposit 351, 1:3 in H2O

Bottom clean Oxygen Plasma treatmentpRF = 220 W, pO2 = 85 mbar, t = 45 s

Pt coating e-beam evaporation

Lift-off heat Microposit Remover Solution 1165 (Shipley) to 60°C,immerse wafer, lift-off occurs almost immediately, rub clean with rubber glove, rinse with plenty of acetone, plasma clean

184 A P P E N D I C E S

Fig. A–9. Improving Pt-lift-off by creating undercut photoresist profiles. Standard processing a) is compared to the toluene soaking pro-cess b), resulting in the final Pt patterns c) and d).

A.2.3. Backside alignment for etching

The Suss MA/BA-6 mask aligner is equipped for backside alignment (SussGmbH, Munich, Germany). First, the alignment marks of the photomaskare photographed through a bottom side microscope looking upwards. Thisimage is saved and transferred to a screen. Secondly, the sample wafer isloaded, frontside down, backside facing the mask for exposure. Using againthe bottom side microscope, corresponding alignment marks on the frontside

A P P E N D I C E S 185

of the wafer (pointing down) are sought. This image is overlaid on the samescreen in half-transparency mode, so that the prerecorded positions of thealignment marks from the mask are visible as well. The alignment marks arethen coincided with each other and the exposure is performed.

A.2.4. Backside Deep Reactive Ion Etching

One current trend in MEMS processing to get rid of wet etching steps inorder to improve the quality and cleanness of the substrates. Deeper trenchesinto silicon can also be etched by inductively coupled plasma technology(Deep Reactive Ion Etching or DRIE technologies), not only by wet hydrox-ide etching. Further advantages include the higher etching rates, the possibil-ity to create deep (> 20 µm) structures with perfectly vertical sidewalls andthe easy masking by a layer of photoresist. Using the standard anisotropichydroxide etch (either KOH or isotropic tetramethylammoniumhydroxideTMAH), the side walls are always angled 54.7° corresponding to the <111>surface of the silicon crystal (when using <100> wafers). For the release of thethin-film membranes by etching from the backside, an anisotropic DRIEprocess had become available in the laboratory of the IMES at ETHZ. Thepractical problem of protecting the frontside of the microhotplate wafers dur-ing a 8-9 hr wet hydroxide etching process could elegantly be circumvented.

The etching process used relies on a patented procedure by Bosch GmbH,Stuttgart, Germany.[1] It was also characterized extensively by Ayon et al.[2,3]

Sulfur hexafluoride, SF6, is introduced at low pressure and using RF power, aplasma is ignited. The SF6 plasma contains a significant amount of F atomswhich readily etch silicon, forming SiF4 (g). In order to protect the sidewallsfrom etching, the SF6 plasma is stopped and a mixture of SF6 and C4F8 isintroduced. Again, a plasma is ignited that deposits a Teflon-like film on allSi surfaces. Detailed characterisations of those films could not be found.Changing back to pure SF6 plasma, the protecting film at the bottom sur-faces is removed first, while the sidewalls remain protected a little longer. Ifthe times of both passivating and etching step are correctly tuned, the side-walls are protected through the whole etching process. These steps are cycleduntil the desired etch depth is reached.

186 A P P E N D I C E S

The system used was a Surface Technology Systems STS multiplex ICP(Newport, UK). The parameters for the backside etching experiments aregiven in Table A–3. All wafers were protected by a 6 µm layer of AZ4562resist on the frontside and the same patterned resist mask on the backside.They were glued to a 100 mm supporting wafer using “white wax” for tworeasons. One, wafers inside the plasma chamber need to be cooled which isdone by a stream of He gas at higher pressure flowing to the backside of thewafer. Therefore, it is important to use very clean and flat backsides in orderto seal correctly the between the plasma chamber and the cooling chamber.Before any process, the He leakage rate into the plasma chamber is measuredand processes can only be started if the He leakage remains below a tightthreshold value. Two, the mechanical strength of the wafer is weakened dur-ing DRIE since not only the membranes are released but also snappinggrooves are etched. Wafers must not break during insertion or removal by theautomated STS system. Inside the system, the wafers are pushed and sup-ported temporarily on only 3 pins. Prior to DRIE, the Si3N4 layers wereopened in a standard RIE process tailored for oxide/nitride removal (STSmultiplex RIE, Newport, UK).

Table A–3. STS multiplex ICP parameters for backside etching.

Etch cyclePassivation cycle

Chiller Temp. 15°C

Bias coil power 10 W

Etch pressure 20 mTorr Pass. pressure 9 mTorr

Etch time 16 s Pass. time 6 s

Etch flow C4F8 0 sccm Pass. RF power 800 W

Etch flow SF6 130 sccm Pass. flow C4F8 100 sccm

Etch flow O2 13 sccm Pass. flow SF6 0 sccm

Etch Flow Ar 0 sccm Pass. flow O2 0 sccm

Etch RF power 1000 W Pass. Flow Ar 0 sccm

A P P E N D I C E S 187

As the etching rate usually varies in different runs, an arbitrary number ofcycles was run first, then the wafer was removed and the resulting trenchdepth measured. According to the etching rate obtained from this measure-ment, the number of subsequent cycles needed to cut through the remainingdepth was determined as summarized in Table A–4. It was important not tooveretch for too many cycles, or the membranes would break during thetreatment. Silicon nitride also etches under these conditions, although at amuch slower rate of ca. 1% with respect to Si.

A.2.5. Setup for Gas Sensitivity Measurements

A small facility for accurate mixing and injecting of binary gas mixtures wasbuilt. Two mass flow controllers El-Flow Digital were obtained fromBronkhorst (Ruurlo, Netherlands). One was calibrated for 0-300 sccm air,the other for 0-150 sccm N2. Featuring a digital interface with bidirectionalcommunication over a common bus system, it was possible to control thestates of all controllers from a computer and to read back the measured valuesof the actual gas flows. All other devices (multiplexer switch and multimeters)were controlled and read out over the GPIB bus (IEEE 488.2). A schematicof the layout is given in Fig. A–10. The on-chip heaters were driven either bya standard DC power supply for the sapphire-based versions or by simple

Table A–4. DRIE number of cycles used for backside etching and corresponding measured etching rates in µm min-1. Labels “mh135” etc. denote the experimental number assigned to individual wafers.

Run No. mh135 mh147 mh157 mh158

1, cycles 90 150 150 290

rate / µm min-1 3.2 4.18 3.4 3.2

2, cycles 332 150 204 77

rate / µm min-1 4.18 2.7 3.2

3, cycles 9 68 3

4, cycles 5

188 A P P E N D I C E S

batteries for the microhotplate versions. The Lakeshore LS 331 digital PID-temperature controller was first used to power the hotplates. However, micro-hotplate membranes kept breaking after only few minutes on power forunknown reasons, destroying the sensor completely. It was then decided toconnect a large capacitor in parallel, to set the current manually by a potenti-ometer and to use only batteries in order to deliver a good quality, distur-bance free DC current. The layout of this small circuit is given in Fig. A–11.

The following parameters could be recorded in gas sensitivity measure-ments: time, tin oxide resistance, flow state of the mass flow controllers, cur-rent on heater, voltage on heater and RTD resistance. These data werecollected by Labview and continuously written into a text file for later analy-sis. The switch system was programmed manually. Therefore, sorting thedata for each channel was performed by the data analysis software Igor Pro 4.

Fig. A–10. Setup for gas sensitivity measurements.

MultiplexerSwitch K7001

Sensor resistanceK2001/K617

MicroSensors

MFC

MFC

Air, 40% RH

Analyte Gas: 2000 ppm / N2

PC: Labview

GPIB/IEEE 488

serial / Flow-Bus

Test chamber

HeaterBattery /Power Supply

V

I

RTD resistanceLS 331

RTD

SnO2 channels

electrical wiring

A P P E N D I C E S 189

Fig. A–11. Dampening circuit. Batteries of 4.5 or 7.0 V were used to deliver the DC current. A capacitor was connected in order to dampen eventual spikes during power-up/down.

A.3. Thermal Characterisation of Microhotplates

A.3.1. Temperature Calibration

The difference in resistance of the RTD temperature sensor on the secondgeneration of microhotplates correlates linearly with increasing heating power(Fig. A–12a). However, the difficulty lies in the step to derive an actual tem-perature value. Even if the RTD was made of Pt with very well defined posi-tive temperature coefficient of 0.0035 Ω(ΩK)-1, there are differences whensputtering or evaporating thin films.

2 Ω -10 kΩ

Pt heater on chip, 420 Ω4.5 V /

7.0 V

6.8 µF

190 A P P E N D I C E S

Fig. A–12. Temperature calibration on microhotplates. a) heating power vs. RTD difference to ambient T signal. b) heating power vs. observed melting points of organic crystals.

Without a four-point geometry on the hot area of the membrane, the resis-tances of the wiring could not be separated from the resistance of the heatedPt region of interest. It was estimated that the unheated Pt thin films adjacent

300

250

200

150

100

50

0

T m /˚

C

50403020100Heating Power /mW

p-anisic acid185˚C

caffeine235˚C

anthra-quinone285˚C

40

30

20

10

0

RTD

sig

nal

dif

fere

nce

to

am

bie

nt

/Ω

50403020100Heating power /mW

RTD =1.27 + 0.64*p(mW)

CentralRTD

a

b

A P P E N D I C E S 191

to the membrane had the largest influence. The simple experiment, heatingthe whole chip in an oven and observing the RTD resistance, failed due tothese unseparable resistances.

Therefore, the method of observing the melting of small grains with wellknown melting temperatures was chosen. Organic crystals are well suited forthis, since the remainders can be washed away with acetone or hexane afterthe observed melting and the hotplate chip can be reused several times. Asshown in Fig. A–13 with the example of caffeine with Tm = 235°C, thepower was recorded, when the small grains marked with arrows melted.Often, the evaporated material recondensed just outside the hot area, form-ing a halo around the membrane. It nicely illustrates in a visual fashion thatthe outside region of the membrane remains at low temperature. However,the obtained calibration (Fig. A–12b) shows a large error. Often, the observa-tions are ambiguous, it was not clear at which point of melting the power wasto be recorded. The data shown are based on observing the melting of thefirst two small grains at two seperate locations on the membrane.

Fig. A–13. Caffeine grains on the membrane a) before and b) after heating to melting temperature.

192 A P P E N D I C E S

With that calibration, a ∆RTD coefficient of approx. 10 °/Ω can be esti-mated. In all subsequent experiments, this coefficient and the RTD signalwere used for membrane temperatures determination.

A.3.2. Heat losses of Microhotplates

Calculations of energy consumption of microhotplates have been discussedin a recent review.[4] When the heating power is switched on, the tempera-ture on the membrane increases until the equilibrium between heat lossesand the heating power is reached. Heat losses consist of the sum of heat con-duction within the membrane Qmem and in air Qair, convection Qcon andradiation Qrad, see Fig. A–14.

(1)

These contribution can be broken down in a geometry factor G, the thermalconductivities λ and the temperature difference heated area to ambient, T -Tamb. In the radiation term, σ is the Boltzmann constant, ε the emissivity ofthe material. With ∆x, all other contributions such as convection are repre-sented. For the evaluation of heat conduction in the membrane it is easier toassume a round cylindrical membrane in cylindrical coordinates with radii ri

for the actually heated central portion of the membrane and ra for the outermembrane diameter at room temperature. In this case, heat conduction onlyhas a radial component.

(2)

with λm being the thermal conductivity of the membrane material, d thethickness, Tamb ambient temperature. Likewise, heat conduction in air ontop of the membrane is

, (3)

Q Qmem Qair Qcon Qrad

Gmemλmem T Tamb–( ) Gairλair T Tamb–( ) Gmemλmem T Tamb–( )Gradσε T

4Tamb

4–( ) ∆x

+ +

+ +

=

+ + +=

Qmem

2πλmd T Tamb–( )ra ri⁄( )ln

-------------------------------------------=

Qair1

4πλair T Tamb–( )1 ri 1 ra⁄–⁄

------------------------------------------=

A P P E N D I C E S 193

assuming a spherical heat flow and that convection is negligible which isoften the case for microstructures. The situation below the membrane is geo-metrically a little different. Height h of the cavity below the membrane is theimportant factor in the geometry term.

(4)

For the radiation, Grad equals simply the hot area of twice a circle with radiusri, A.

(5)

The membrane should be designed as thin as possible and the ratio ri/ra ashigh as possible, e.g. 3, for minimizing the heat losses in microhotplates.

Fig. A–14. a) heat losses from the heated center of the micromachined membrane. b) equivalent hotplate in cylindrical coordinates for easier calculation of heat conduction within the membrane. For details see text.

Qair2

Aλair T Tamb–( )h

---------------------------------------=

Q 2Aσε T4

Tamb4

–( )=

ri

ra

a b

Qair + Qconv. + Qrad.

Qmembrane

194 A P P E N D I C E S

A.3.3. Numerical Evaluation

In this section, numerical values for the microhotplate design developed inthis work are calculated, see Table A–5. The temperature difference is set to300K. Simon et al. collected most of the relevant material constants,[4] othervalues were taken from the CRC Handbook of Physics and Chemistry[5] andare marked with asterisks. The experimentally obtained heating power for sustaining a temperature of320°C is approx. 50 mW. The sum of all heat losses of 147 mW is thereforetoo high but in the same range. Qair1 = 109 mW seems to be the most out ofplace result. Whereas in (3) only the heat conductance of air appears, theproperties of heat transfer from the hot membrane to air are not considered.Especially when it comes to determining the heat transfer properties of mate-rials and thin films, values can be extremely difficult to measure. Exact exper-imental thermal characterisations are not often published. However, the useof FEM simulations is widespread, e.g. see Briand et al.[6,7] Often, a combi-nation of simulation and measured temperatures from real hotplates are usedto derive numeric values which should be treated cautiously. Simon et al. listtwo examples of an estimation of the same heat transfer coefficient frommembrane to air αm = 125 Wm-2K-1 and αm = 30 Wm-2K-1. One shouldalso carefully observe the model assumptions and boundary conditions thatwere used. However, if a Qair_sim is calculated with the geometries from thepresent design (Table A–5),

(6)

a more realistic value of heat loss of 24 mW results for αm = 125 Wm-2K-1. Amodified accounting with air losses calculated according to these simulationsis given in Table A–6. The sum of 62.2 mW is also close to the experiment.

Radiation losses are not very pronounced at ∆T = 300 K, only 2.9% of alllosses. Due to the T4 dependence, the radiation losses will quickly becomesignificant when going to higher temperatures. This leads to a deviation fromlinear behaviour of a temperature vs. heating power curve, see also Fig. 1–7in Chapter 1.

Qairsim Aαm T Tamb–( )=

A P P E N D I C E S 195

In conclusion, the heating power to sustain a temperature of 320°C is dissi-pated by two almost equally important pathways, heat conduction within themembrane and loss to the air. Radiation losses do not play an important roleat low temperatures.

Table A–5. Accounting of the heat losses in the designed microhotplates.

Parameter Value Unit Loss /mW %

Membrane lossesa

Qmem, Eq. (2)

a. Losses from the Pt electrode/heater layer not considered.

λSi3N4 30 W/mK b

b. λSi3N4 is a function of temperature and also varies with the pro-cessing of silicon nitride, range 9-30 W/mK.

d 2.5 · 10-7 m

T - Tamb 300 K

ra 0.45 · 10-3 m c

c. Radii calculated as a circle with same area as designed.

ri 0.85 · 10-3 m 22.5 15.4

Air losses frontsideQair1, Eq. (3)

λair 0.03 W/mK d

d. λair varies from 0.026 to 0.044 W/mK at T = 20°C, resp. T = 320°C.

108.1 74.2

Air losses backsideQair2, Eq. (4)

A 6.4 · 10-7 m2

h 0.525 · 10-3 m e

e. Silicon wafer thickness 525 µm.

10.9 7.5

Radiation lossesQrad, Eq. (5)

σ 5.67 · 10-8 J K-4m-2s-1

ε 0.5

T 593.15 K

Tamb 298.15 K 4.2 2.9

Total losses 145.7 100

196 A P P E N D I C E S

A.3.4. Heating Kinetics of Microhotplates

The time-dependent heat balance can be expressed in analogy to electricalcircuits as

. (7)

A thermal capacitance C and a thermal Resistance Rtherm are assigned to thehotplate. These are characteristic for each individual design. Solving this dif-ferential equation for T(t) yields the following expression

(8)

with the characteristic time constant τ. Since T(t) and Pel can be measured,all the characteristic constants of the hotplate can be calculated.

Table A–6. Accounting of the heat losses using heat transfer values from simulation work.

Parameter Value Unit Loss /mW %

Membrane lossesQmem, Eq. (2)

22.5 36.2

Air losses frontsideQairsim1, Eq. (6)

αairsim1 125 W/m2K 24.0 38.6

Air losses backsideQairsim2, Eq. (6)

aairsim2 60 W/m2K 11.5 18.5

Radiation lossesQrad, Eq. (5)

4.2 6.7

Total losses 62.2 100

Pel CdT t( )

dt-------------

T t( ) Tamb–

Rtherm

----------------------------+=

T t( ) Tamb PelRtherm 1 et τ⁄–

–( )τ RthermC=

=–

A P P E N D I C E S 197

Fig. A–15. Electrical wiring for measuring the time response of the micro-hotplate upon heating and cooling down.

A.3.5. Experimental

The electrical response of heating up to approx. 350°C (4.5 V) and coolingdown to ambient was recorded using a LeCroy 9450 digital oscilloscope. Theelectrical setup is shown in Fig. A–15. The oscilloscope was connected inparallel with the Lakeshore 331 controller that recorded the resistance of theRTD strip. A 4.5 V battery was directly connected to the heater. Its switchingtime was also measured using a 470 Ω foil resistor: < 1 µs. All data was col-lected using LabView.

A.3.6. Results and Discussion of Kinetics

The resulting build-up and decay-curves are given in Fig. A–16. The noisewas reduced by applying moving average filters, then exponential curves werefitted to the data. Evaluating Eq. (8) with the data yields a time constantτ = 4.02 ms. Rtherm can be calculated by applying a steady state condition to(7), dT/dt = 0. It can be rearranged to give

(9)

At approx. 50 mW for 300°C temperature difference, Rtherm = 6 Ks J-1 andthe thermal capacitance 6.7 · 10-4 J K-1.

Pt heater on chip, 470 Ω

4.5 V RTDstrip305 Ω

LS 331 in4pt mode

LeCroy9450 Osc.

Rtherm

T t( ) Tamb–

Pel

----------------------------=

198 A P P E N D I C E S

Fig. A–16. a) RTD response for heating up to 4.5 V. b) cooling down after power is switched off. For both cases, the trigger was set to +15 ms where the build-up, resp. the decay of the signal begins.

The time constant for cooling from 350°C to ambient temperature is evenshorter, τ = 2.5 ms, i.e. the microhotplate is at ambient temperature within8 ms (approx. 3τ) from cutting the heating power.Due to the only 250 nm thin membrane, the microhotplates fabricated havea time constant of 4 ms which is very fast, even compared to other designspublished. 350°C can be reached within only 15 ms (approx. 3τ) which is thebasis for enhanced sensing schemes like temperature pulsing or energy savingby switching the sensor heater off when not in use.

a

b

Vo

ltag

e /a

.u.

50ms403020100time

signal exp. fit

Vo

ltag

e /a

.u.

50ms403020100time

signal exp. fit

A P P E N D I C E S 199

A.4. Powder Characteristics

Powder size distributions were measured using a centrifuge method(Brookhaven BI-XDC, Brookhaven Instruments Corp., Holtsville, NY, USA)where the x-ray absorbance as a function of powder sedimentation could beobserved. Particle diameters then can be calculated using the Stokes law.

A.4.1. Cerac Tin Oxide Powder

In ESA experiments assessing the surface charge at varying pH values, an iso-electric point of 3.7 was obtained (Fig. A–17). In order to stabilize colloidaldispersions of SnO2, the pH was set to 8 for maximum negative surfacecharge.

Table A–7. Size distribution of Cerac SnO2 as obtained.

d50 224 nm

d10 144 nm

d90 313 nm

Volumetric mean 225 nm

Calc. Specific Surface 4.3 m2/g

x-ra

y Si

gn

al /a

.u.

4 5 6 7 8 90.1

2 3 4 5 6 7 8 91

Particle Diameter / µm

100

80

60

40

20

0

Cu

mu

lative Mass/%

SnO2 Cerac, Lot 23339, Producer: 0.22 µm

200 A P P E N D I C E S

Fig. A–17. Electrosonic Kinetic Amplitude (ESA) measurement of tin oxide surface charges.

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

ESA

Sig

nal

/ m

Pa

m V

-1

1098765432pH

A P P E N D I C E S 201

A.4.2. Alfa Aesar Tin Oxide Nanometer Dispersion

The dispersion was 15 wt% in water, pH =10.1. The AgNO3 test with super-natant solution of the dispersion indicated minute amounts of Cl-. DTA/TGA measurements showed no indication of organic additives.

A.4.3. Reference Gas Sensitivity of the Tin Oxide Powders

The SnO2 powders used for this work were coated onto standard thick-filmsensor substrates from IPTC. These are small alumina thick-film strips withinterdigitated Au electrodes on the front and heater pattern on the back side.These substrates are all screen-printed, cheap and serve the IPTC as workinghorse for assessing the gas sensing properties of powders.

Table A–8. Size distribution Alfa Aesar tin oxide.

d50 10.3 nm

d10 7.6 nm

d90 13.0 nm

Volumetric mean 10.1 nm

Calc. Specific Surface 90 m2/g

x-ra

y Si

gn

al /a

.u.

5 6 7 8 90.01

2 3 4 5 6 7 8 90.1

Particle Diameter / µm

100

80

60

40

20

0

Cu

mu

lative Mass/%

SnO2 Alfa Aesar, nanoscaled colloid

Producer: 0.015 µm

202 A P P E N D I C E S

A few droplets of 25 vol% dispersion of 220 nm SnO2 powder from Ceracwas coated over the electrodes in a humid environment for slow drying.Annealing was performed at 800°C/5 hrs. The nano-scaled SnO2 dispersionwas used as received for drop-coating. However, no annealing was performedin order to have a similar treatment as the MIMIC SnO2 lines on microhot-plates.

The responses for CO and C2H5OH at gas concentrations of 250 to500 ppm in wet and dry air are shown in Fig. A–18. Obviously, CO at250 ppm could not be detected, whereas pure tin oxide shows good sensitiv-ity to 250 ppm C2H5OH. The nano-scaled droplet-coated film exhibited thehigher resistance in the range of 100 MΩ, but also a higher response. Theresistance drops over more than two orders of magnitude. This is in goodagreement with the argument that nano-scaled powders offer improved sensi-tivities.

In summary, the limiting sensitivity of approx. 500-600 ppm to COobserved with microline sensors is not a shortcoming of the miniaturizeddesign, it is due to the limited sensitivity of the pure tin oxide that was usedas gas sensing material.

Fig. A–18. Gas sensitivity measurements using the pure tin oxide powders.

102

103

104

105

106

107

108

109

Res

ista

nce

/Ω

8006004002000

Time /min

0

250

500

CO

, EtOH

/pp

m

Alfa Aesar Tin Oxide 10 nm Cerac Tin Oxide 220 nm

CO Et. CO Et. CO Et. CO Et. CO Et. CO

dry air

A P P E N D I C E S 203

A resistance vs. temperature curve of pure tin oxide was measured early in thethesis work by tape-casting the 33 vol% dispersion of Cerac powder onto asaphhire wafer and firing the layer at 800°C/5 hrs. Electrical contacts werepainted with Pt-colloid over which Pt wires were glued and fixed using aninorganic glue. The contacts were annealed at 800°C before measurementwhich was carried out in a tube furnace equipped with a thermocouple fortemperature measurement.

The n-type semiconducting behaviour is given in Fig. A–19. Theobtained value for the bandgap energy of SnO2 of 3.56 eV is in good agree-ment with literature values.

Fig. A–19. Tin oxide resistance in air (Cerac powder). From the Arrhenius plot in the inset, a band gap of 3.56 eV was obtained.

160

140

120

100

80

60

40

20

0

Tin

Oxi

de

Res

ista

nce

/MΩ

600550500450Temperature /˚C

-13

-12

-11

-10

-9

-8

-7

ln( m

Ω)

1.401.301.201.101000 K / T

σ

204 A P P E N D I C E S

A.5. ToF-SIMS Spectra of Microcontact-Printed Surfaces

A.5.1. ToF-SIMS

ToF-SIMS (Time-of-Flight Secondary Ion Mass Spectrometry) is anextremely sensitive analytical method to obtain mass spectra of moleculespresent on surfaces. The sensitivity is sufficient to probe monolayer cover-ages. A primary ion beam of Cs+ or Ar+ ions is accelerated with keV energiestowards the sample surface. At impact of primary ions, the analyte moleculesare desorbed, fragmented and partially ionized. The charged ions can then beextracted into a Time-of-Flight mass analyzer. If the ion dosage of the pri-mary beam is augmented to values above 1012-1013 ions/cm2, the beamcauses etching of the surface. Thus, depth profiling at molecular resolution ispossible which is referred to as dynamic SIMS. At the same time, lateral reso-lutions of approx. 20 µm can be achieved if the ion beam is focussed. Pat-terned surfaces can be imaged by rastering the beam across the surface.However, quantitative results can only be obtained by careful concentrationstandardization due to the fact that ionization efficiency is strongly depen-dent on the chemical nature of the samples. For more information see[8] andreferences therein. All experiments described were carried out in static SIMSmode.

A.5.2. Thiol-gold SAM layers

In the positive Spectra, the typical pattern m/z = 27/29, 39/41/43, 53/55/57,67/69/71 of alkanes is observed.[9] However, these may also originate fromresidual organics and are not suited for the identification of SAM-layers.191 Cs+ (primary ion beam) and Au+ (surface) are prominently present aswell. Since alkanes do not form many stable anions whereas Au and Sulfur-containing species do, the negative mass spectra yield more information andare less difficult to interpret. Literature reference ToF-SIMS data of alkaneth-iol SAMs on Au including an extensive list of possible fragments was avail-able.[10] In the case of PDMS reference layers, the series of PDMS oligomers74 (Si(CH3)2O), 132 (Si2(CH3)4O), 148 (2 Si(CH3)2O), 206(Si3(CH3)6O2) and 207 (Si3(CH3)5O3) could be found in positive modemass spectra.

A P P E N D I C E S 205

Two spectra of HDT-Au SAM samples are compared in Fig. A–20. One (a) isthe negative mode mass spectrum of the microcontact-printed sample with aPDMS mold contact time of 100 s, the other spectrum (b) was obtainedfrom the SAM reference that was prepared by immersion in HDT/ethanolsolution. The spectra match very well, there are no obvious differences. Thefragments that could be identified are listed in Table A–9. The formation ofvarious AuS-fragments proved very useful for direct confirmation of a sulfur-gold chemical bond. If we designate HDT as the molecular ion M, the frag-ment (M-H)- could also be found albeit as a very weak signal. Other impor-tant fragments include AuSCH2

- and AuSCH2CH2- which again proves the

bonding of an alkane to S and Au. In the low mass range, the peaks for Si-containing species were also found at low intensities. The di- and tri-mono-meric units of PDMS at m/z = 119, 133, 149, 223 were barely present at justa few counts above noise level (not resolved in Fig. A–20).

206 A P P E N D I C E S

Fig. A–20. a) ToF-SIMS negative mode mass spectra with various magnifications of microcontact-printed gold surfaces with hexadecanethiol. b) reference mass spectra of a SAM-coated gold surface from a HDT/ethanol solution.

0 100 200 300 400 5000

2

4

6

8x 10

4

m/z

Co

un

ts

MH15507.SUR

Co

un

ts

0 100 200 300 400 5000

2

4

6

8x 10

4

m/z

MH15509.SUR

0 10 20 30 40 50 60 70 80 90 1000

2

4

6

8x 10

4

m/z

MH15509.SUR

190 200 210 220 230 240 250 260 270 280 2900

2000

4000

6000

8000

m/z

Co

un

ts

MH15509.SUR

390 395 400 405 410 415 420 425 4300

100

200

300

m/z

MH15509.SUR

190 200 210 220 230 240 250 260 270 280 2900

2000

4000

6000

8000

m/z

Co

un

ts

MH15507.SUR

Au-

196.97AuS

-

228.94AuSCH2CH2

-

256.97

AuS2-

260.91

390 395 400 405 410 415 420 425 4300

100

200

300

m/z

MH15507.SUR

Au2-

393.93Au2S

-

425.91

a

b

0 10 20 30 40 50 60 70 80 90 1000

2

4

6

8x 10

4

m/z

MH15507.SUR

CH-

13C2H

-

25O2

-/ S

-

32

SiO2-

59.97HSiO3

-

76.97

A P P E N D I C E S 207

A.5.3. Microcontact-Printed Octadecyltrichlorosilane Layers

Microcontact-printed OTS layers on Si are not as easily identified as in thecase of Au-S SAMs. A brief overview on the elements involved makes thispoint obvious. The layer is of similar composition as the substrate. Si possiblyoriginates from the wafer, its native SiO2 layer, from PDMS and from OTS.Cl is expected to evaporate as HCl during the coating process and generalalkane signals are too unspecific for identification of OTS. Therefore, theonly mean of unambiguously establishing the presence of an OTS-layer is the

Table A–9. List of identified anionic fragments in Tof-SIMS spectra ordered by mass and type of layer

a) low mass anions b) higher mass anions

m/z Mass /u Ion m/z Mass /u Ion

Silicon/PDMS/OTS fragments, unspecific

HDT SAM on Au fragments

12 12.0004 C- 197 196.9665 Au-

13 13.0082 CH- 229 228.9386 AuS-

14 14.0163 CH2- 243 242.9542 AuSCH2

-

16 15.9934 O- 257 256.9699 AuSCH2CH2-

17 17.0021 HO- 257a

a. Two fragments with the same nominal mass could be resolved well.

257.2302 M-H = S(CH2)15CH3

-

24 24.0003 C2- 261 260.9107 AuS2

-

25 25.0081 C2H- 394 393.9331 Au2-

32 31.9887 O2- 422 422.2247 Au(CH2)15CH3

-

35 34.968 Cl- 426 425.9052 Au2S-

41 41.0034 C2HO- 454 454.1968 Au(M-H)-

59 59.0141 C2H3O2- 711 711.4271 Au(M-H)2

-

60 59.9656 SiO2-

77 76.9688 HSiO3-

208 A P P E N D I C E S

detection of intact octadecyltrioxysilane species, e.g. SiO3(CH2)17CH3-.

Such fragments have also been identified by Houssiau et al. examining OTSSAM on aluminum metal surfaces.[11] The presence of octadecyl or relatedspecies may provide additional hints.

Fig. A–21. ToF-SIMS negative mode mass spectra of microcontact-printed OTS-samples and references, one coated in OTS solution, the other by print-ing several times with a dry PDMS mold.

327 328 329 330 331 332 333 334 335 336 3370

50

100

150

200

250

m/z

Co

un

ts

MH15533C.SUR

327 328 329 330 331 332 333 334 335 336 3370

5

10

15

20

25

m/z

Co

un

ts

MH19010C.SUR

b) Comparing to OTS coating

c) PDMS fragments not present on OTS reference d) Resolving different ions of same m/z

100 120 140 160 180 200 220 2400

1000

2000

3000

m/z

Co

un

ts

MH19006C.SUR

100 120 140 160 180 200 220 2400

200

400

600

800

1000

m/z

Co

un

ts

MH19010C.SUR

µ-CP sample

PDMS reference

218 219 220 221 222 223 224 225 226 227 2280

1

2

3

4

5

6

m/z

Co

un

ts

MH15533C.SUR

218 219 220 221 222 223 224 225 226 227 2280

50

100

150

200

250

m/z

Co

un

ts

MH19010C.SUR

118.6 118.8 119 119.2 119.40

5

10

15

20

25

m/z

Co

un

ts

MH15533C.SUR

118.6 118.8 119 119.2 119.40

20

40

60

80

m/z

Co

un

ts

MH19010C.SUR

a) Comparing to PDMS coating

µ-CP sample

OTS reference

µ-CP sample

OTS reference

µ-CP sample

OTS reference

m/∆m = 4100

A P P E N D I C E S 209

The resulting negative mode mass spectra of OTS-printed samples are shownin Fig. A–21. Part a) only shows the more specific m/z range above 100. Thespectrum of the microcontact-printed sample strongly resembles the PDMSreference spectrum. The di- and trimonomeric fragment series m/z = 119(CH3Si2O3), 133 (C2H5Si2O3), 149 (C3H9Si2O3), 223 (C5H15Si3O4) areprominently present including Si isotope signals and maybe some fragmentslacking H-atoms. These fragments together with fragments that were identi-fied as OTS indicators are listed in Table A–10. Part b) then shows that OTSfragments like the most intense H2SiO3(CH2)17CH3

- (m/z = 331) alsocould be detected in the microcontact-printed samples, although at roughlyten times lower intensity. To reconfirm that peaks attributed to PDMS donot show up in the OTS reference, the region at m/z = 223 is shown in c).The achieved resolution is assessed in d), where the PDMS-attributed peak atm/z = 119 nominally overlaps with another ion 119 of the OTS reference.However, these fragments do not have the same exact mass and are thereforeseparated. A mass resolution of m/∆m = 4100 was obtained.

Table A–10.List of identified anionic fragments from OTS and PDMS in Tof-SIMS spectra.

m/z Mass /u Ion m/z Mass /u Ion

OTS-specific anions PDMS-specific anionsa

a. The PDMS fragments with m/z ± 16 (oxygen atom) or m/z ± 15 (methyl group) are encountered frequently.

313 313.2563 SiO2(CH2)17CH3- 119 118.9581 CH3Si2O3

-

329 329.2512 M-

SiO3(CH2)17CH3-

133 132.9745 C2H5Si2O3-

330 330.2590 (M+H)-; isotopic

29Si149 149.0068 C3H9Si2O3

-

331 331.2668 (M+2H)-; isotopic

30Si223 223.0244 C5H15Si3O4

-

210 A P P E N D I C E S

A.5.4. Summary of ToF-SIMS analysis

ToF-SIMS analysis confirmed the presence of a thiol-gold SAM with onlyminor amounts of PDMS contamination. In the case of transferring octade-cyltrichlorosilanes by micro-contact printing, there seems to be a consider-able amount of PDMS residues present. The relative amount of detectedOTS fragments decreased by approximately 10 times in, compared to a refer-ence surface prepared by immersion in OTS/hexane.

A P P E N D I C E S 211

A.6. List of Abbreviations

µCP Micro-Contact Printingµ-TAS Micro Total AnalysisµTM Micro Transfer Molding3DP Three-Dimensional Printing Process AFM Atomic Force MicroscopyCMOS Complementary Metal Oxide SemiconductorDCM Direct Ceramic MachiningDMD Digital Micromirror DeviceDRAM Dynamic Random Access MemoryDRIE Deep Reactive Ion EtchingELISA Enzyme-Linked Immunosorbent AssayEPR Electron-Paramagnetic ResonanceGT Green TapeHDT HexadecanethiolLIGA Lithography, Galvanoformung, Abformung (German)LTTC Low-Temperature Co-Fired Ceramic Multilayer TechnologyMALDI Matrix-Assisted Laser Desorption and IonisationMEMS Microelectromechanical SystemsMIMIC Micromolding in CapillariesMOSFET Field Effect Transistor with Metal Oxide Semiconductor gateOTS OctadecanethiolPCR Polymerase Chain ReactionPDMS Poly-DimethylsiloxanePMMA Poly-MethylmethacrylatePRC Photoresist CastingPZT Lead-Zirconia TitanateQMB Quartz MicrobalanceRTD Resistive Temperature DeviceSAM Self-assembled MonolayerSAMIM Solvent-Assisted MicromoldingSAW Surface-Acoustic WaveSEM Scanning Electron MicroscopySTM Scanning Tunneling Microscopy

212 A P P E N D I C E S

TEM Transmission Electron MicroscopyTEOS TetraethoxysilaneToF-SIMS Time-of-Flight Secondary Ion Mass SpectroscopyTPD Temperature Programmed DesorptionVLSI Very Large Scale Integrated CircuitsXPS X-Ray Photoelectron Spectroscopy

A P P E N D I C E S 213

A.7. References

[1] F. Laermer, A. Schilp, Method of anisotropically etching silicon, U.S. Patent, 1996, No. 5,501,893.

[2] A. A. Ayon, R. Braff, C. C. Lin, H. H. Sawin, M. A. Schmidt, Characterization of a time multiplexed inductively coupled plasma etcher, J. Electrochem. Soc., 1999, 146, 1, 339-349.

[3] A. A. Ayon, R. A. Braff, R. Bayt, H. H. Sawin, M. A. Schmidt, Influence of coil power on the etching characteristics in a high density plasma etcher, J. Electrochem. Soc., 1999, 146, 7, 2730-2736.

[4] I. Simon, N. Barsan, M. Bauer, U. Weimar, Micromachined metal oxide gas sen-sors: opportunities to improve sensor performance, Sens. Actuators B, 2001, 73, 1-26.

[5] "CRC Handbook of Chemistry and Physics", 83rd ed., edited by D. R. Lide, CRC Press, Boca Raton, 2001.

[6] D. Briand, A. Krauss, B. van der Schoot, U. Weimar, N. Barsan, W. Gopel, N. F. de Rooij, Design and fabrication of high-temperature micro-hotplates for drop-coated gas sensors, Sens. Actuators B, 2000, 68, 1-3, 223-233.

[7] D. Briand, S. Heimgartner, M. Gretillat, B. van der Schoot, N. F. de Rooij, Ther-mal optimization of micro-hotplates that have a silicon island, J. Micromech. Microeng., 2002, 12, 971.

[8] D. A. Skoog, J. J. Leary, "Principles of Instrumental Analysis", 4th Edition ed., Saunders College Publishing, Orlando, 1992, 8.

[9] F. W. McLafferty, F. Tuecek, "Interpretation of Mass Spectra", 4th ed., University Science Books, Mill Valley, CA, 1993.

[10] D. A. Hutt, G. J. Leggett, Static Secondary Ion Mass Spectrometry studies of self-assembled monolayers: electron beam degradation of alkanethiols on gold, J. Mater. Chem., 1999, 9, 923-928.

[11] L. Houssiau, P. Bertrand, Tof-SIMS study of organosilane self-assembly on alumin-ium surfaces, Appl. Surf. Sc., 2001, 175-176, 351-356.

214 A P P E N D I C E S

215

Curriculum Vitae

Martin Heule Born November 2, 1973Citizen of Zurich and Widnau SG

Education

June 1999 - present

ETH ZurichPhD thesis in the group of Nonmetallic Materials. Supervisor: Prof. Dr. L. J. Gauckler.

May 1999 Dipl. Chem. ETH

Diploma thesis in Analytical Chemistry: “Development of a Near-Field Scanning Microscope to perform Raman Spectroscopy at Liquid-liquid phase boundaries” with Prof. Dr. R. Zenobi, Laboratory of Organic Chemistry.

January 1998 Elective subjects: Analytical Chemistry, Inorganic Chemistry:Practical work"Synthesis and NMR characterisation of RuH(OMe-BIPHEP)(Arene)BF4 Complexes" with Prof. Dr. P. S. Pregosin, Laboratory of Inorganic Chemistry

October 1997 Swiss ArmyLieutenant,Commissioned Officer (Field Artillery), Frauenfeld.

October 1994 ETH ZurichStudies in Chemistry

1989 – 1994 Kantonsschule Zürcher Unterland, BülachMatura Type C

216

Publications

M. Heule, U. P. Schönholzer, L. J. Gauckler, “Patterning Colloidal Suspensions by Selective Wetting of Microcontact-Printed Sur-faces”, submitted to Adv. Mater, 2003.

M. Heule, S. Vuillemin, L. J. Gauckler, “Overview on Powder-based Ceramic Meso- and Microscale Fabrication Processes”, sub-mitted as review article, 2002.

M. Heule, L. Cavalli, L. J. Gauckler, “Miniaturised enzyme reactor based on hierarchically shaped porous ceramic microstruts”,submitted to Adv. Mater, 2002.

M. Heule, L. J. Gauckler, “Miniaturised Arrays of Tin Oxide Gas Sensors on Single Microhotplate Substrates Fabri-cated by Micromolding in Capillaries”, Sens. Actuator B, 2002, (in press).

M. Heule, J. Schell, L. J. Gauckler, “Powder-based Tin Oxide Micro-Components on Silicon Substrates fabricated by Micro-molding in Capillaries”, J. Am. Ceram. Soc, 2002, 85 (6), (in press).

M. Heule, L. J. Gauckler,“Micromachined Nanoparticulate Ceramic Gas Sensor Array On MEMS Substrates”,Mat. Res. Soc. Symp. Proc., 2002, 69-74.

M. Heule, L. J. Gauckler, ”Gas Sensors Fabricated from Ceramic Suspensions by Micromolding in Capillaries”, Adv.Mater. 2001, 13 (23), 1790 -1793.

see also in “Highlights of the Recent Literature / Editors’ Choice”Science, 2002, 295, 15.

M. Heule, L. J. Gauckler, “Microfabrication of Ceramics Based on Colloidal Suspensions and Photoresist Masks”, J.Photopolym. Sci. Technol., 2001, 14 (3), 449.

M. Heule, L. Meier, L. J. Gauckler, “Micropatterning of Ceramics on Substrates towards Gas Sensing Applications”, Mat. Res.Soc. Symp. Proc., 2001, 637, EE 9.4.

217

Conference Presentations

202nd Electrochemical Society Meeting, Salt Lake City, USA,20 – 24 October 2002.“Micro-Ceramic Gas Sensor Arrays” (invited talk).

International Meeting on Chemical Sensors IMCS 9, Boston, MA. USA, 7 – 10 July 2002.“Miniaturized ceramic gas sensor arrays on a single microhotplate fabricated by softlithography” (contributed talk).

7. Arbeitskreistreffen “Ausgangspulver” der Deutschen Keramischen Gesellschaft DKG,EMPA Dübendorf, Switzerland, 23. Mai 2002.“Colloidal Chemistry for Micropatterning of Ceramic Powders” (invited talk).

Materials Research Society (MRS) Fall Meeting 2001, Boston, MA, USA, 26 – 30 November 2001.”Micromachined Nanoparticulate Ceramic Gas Sensor Array On MEMS Substrates”,(contributed talk).

Materials Research Society (MRS) Fall Meeting 2001, Boston, MA, USA, 26 – 30 November 2001,”Transfer of High-Temperature-Annealed Ceramic Microstructures onto Heat-SensitiveSubstrates”, (poster).

Workshop on Advanced Materials, Crêt-Bérard/Puidoux, Switzerland, 9 September 2001,participants from EPFL, ETHZ, UCSB, WIS (materials science, chemistry, physics)“Microfabrication of Ceramics as MEMS Components”, (poster).

12th Conference on Photopolymer Science and Technology, Chiba, Japan, 26 -29 June 2001.“Microfabrication of Ceramics based on Colloidal Suspensions and Photoresist Masks”,(invited talk).

103rd Annual Meeting of the American Ceramic Society, Indianapolis, IN, USA, 23-25 April 2001.”Micro-Ceramics Fabricated by Soft Lithography”, (contributed talk).

Materials Research Society (MRS) Fall Meeting 2000, Boston, MA, USA, 27 November – 1 December 2000.”Micropatterning of Ceramics on Substrates towards Gas Sensing Applications”, (contributed talk).

218